Inductive inference affective language analyzer simulating artificial intelligence

[0002] This application claims priority of provisional application Serial No. 60/157,322 filed Oct. 1, 1999.

[0003] This invention relates to the process of programming the rules governing the use of affective language into a computer, allowing a passive detection mode, along with an active simulation of affective language in an AI mode.

[0004] Language simulation of an ethical nature has remained an elusive goal for researchers in the field due to the complexities involved in simulating affective language in general. Fortunately, such an achievement has finally been achieved with the development of the power pyramid hierarchy (FIGS. 1 through 10). A direct outcome of the power pyramid hierarchy is the formulation of the power pyramid definitions, simulating emotionally-charged language in a form programmable into a computer. Through the aid of these schematic definitions, the logical consequences of the power pyramid hierarchy are programmed directly into a computer, serving in applications calling for the rapid decoding of motivational parameters. The computer further utilizes these decoded results to produce a language simulation of its own, allowing for an ethically based simulation of artificial intelligence.

[0005] The concept of artificial intelligence (AI) refers to language simulated using a computer. The prevailing standard in the field is the deductive inference machine, which employs deductive reasoning to establish original conclusions from a standard battery of logical premises. The product of years of research by ICOT (The Institute of New Generation Computer Technology) the deductive-inference machine uses information stored in a regional database to deductively draw fresh conclusions not literally contained in the data. A major drawback to this deductive format, however, is a basic restriction limiting conclusions to premises immediately at hand.

[0006] An alternate form of reasoning (known as induction) is much better suited to an ethical simulation of AI, formally gathering together the best available evidence, in order to draw the most probable conclusion from the sum total of facts. In contrast to deductive reasoning, the conclusions derived through inductive reasoning are never absolutely certain, although the uncertainties of the physical world give inductive reasoning the clear advantage in such a problem-solving mode. According to such an inductive model, each individual builds up a mental model of reality over a lifetime, forming a master template for all current experiences. When expectations match surroundings, a general sense of security is reached. A mismatch leads to a surprised reaction, followed by investigative behavior. An ethically based AI computer is similarly equipped with its own formal map of reality, employed in an analogous detection and matching function. The logistics of the power pyramid hierarchy rightfully enter the picture at this juncture, serving as the foundation for the first inductive system for decoding and simulating affective language. The logical attributes of the power pyramid definitions provide a formal model of motivational behavior, allowing a decoding of the motivational parameters of a given verbal interchange. On the basis of this determination, the computer devises counter-response of its own, effectively simulating a sense of motivation in the verbal interaction.

[0007] The Inductive Inference Affective Language Analyzer (hereafter abbreviated IIALA) makes extensive use of parallel processing, where various aspects of a complex problem are handled simultaneously, minimizing the computational bottleneck plaguing sequential processing. The number of parallel processor complexes equals the sum-total of power pyramid definitions (for a grand total of 320), a feasible number even by today's design standards. These processor complexes are further organized in a hierarchial fashion, mirroring the stepwise architecture of the power pyramid hierarchy. This hierarchial arrangement takes full-advantage of the strict transformational logic of the power pyramid hierarchy, eliminating much of the redundancy certain to occur in any convincing language simulation. The greatest degree of complexity involves programming at the most basic (personal) levels of the power pyramid hierarchy, the remaining higher authority levels extending naturally upon this elementary foundation.

[0008] The most appropriate unit of input is necessarily the sentence, for the power pyramid definitions are analogously organized in the form of a dual sentence structure. With a design schematic specifying a parallel array of 320 dedicated processor complexes, each individual complex employs parallel processing to determine the relative degree of correlation between an input (target) sentence and its matching power pyramid definition template. The matching process scrutinizes the various grammatical elements of a given sentence, statistical correlating these specifics to a given power pyramid definition. The verb tense, the plurality and person of the nouns/pronouns etc. are all scrutinized according to pre-set criteria. Each processor complex then calculates the sum-total of correct matches, yielding the relative probability that a given sentence matches a particular power pyramid definition. The processor complex yielding the highest overall rating is singled out as the most probable solution to the power pyramid definition matching procedure.

[0009] The context of the interaction is further taken into account through the aid of a feedback circuit, the priority of the individual microprocessor complexes preemptively weighted on the basis of preceding deliberations. Each power pyramid definition is composed of both past (as well as present) design components, establishing context as a further critical feature in the detection procedure. The IIALA retains in long-term storage a record of every relevant experience with a given individual or situation. On this contextual basis, the master control unit selectively weights the individual processor complexes according to a record of both past (as well as current) behavior patterns. In this respect, the IIALA is exquisitely sensitive to variations in personality Oust as humans are instinctively so) satisfying yet a further prerequisite of Turing's test (the ultimate standard for evaluating such a system).

[0010] FIG. 1A is virtuous mode of the power pyramid hierarchy

[0011] FIGS. 1B, 1C, 1D and 1E are the power pyramid definitions for the virtuous mode

[0012] FIGS. 2(A) and (B) are the two-stage, time-based schematic of operant conditioning

[0013] FIG. 3A is the vices of defect of the power pyramid hierarchy

[0014] FIGS. 3B, 3C, 3D and 3E are the power pyramid definitions for the vices of defect

[0015] FIG. 4A is the accessory virtues, values, and ideals of the power pyramid hierarchy

[0016] FIGS. 4B, 4C, 4D and 4E are the power pyramid definitions for the accessory virtuous mode

[0017] FIG. 5A is the accessory vices of defect of the power pyramid hierarchy

[0018] FIGS. 5B, 5C, 5D and 5E are the power pyramid definitions for the accessory vices of defect

[0019] FIG. 6A is the vices of excess of the power pyramid hierarchy

[0020] FIGS. 6B, 6C, 6D and 6E are the power pyramid definitions for the vices of excess

[0021] FIG. 7A is the accessory vices of excess of the power pyramid hierarchy

[0022] FIGS. 7B, 7C, 7D and 7E are the power pyramid definitions for the accessory vices of excess

[0023] FIG. 8A is the groupings of hyperviolence of the power pyramid hierarchy

[0024] FIGS. 8B, 8C, 8D and 8E are the power pyramid definitions for the terms of hyperviolence

[0025] FIG. 9A is the accessory groupings of hyperviolence of the power pyramid hierarchy

[0026] FIGS. 9B, 9C, 9D and 9E are the power pyramid definitions for accessory terms of hyperviolence

[0027] FIG. 10 is the master diagram of the power pyramid schematic illustrating the 5 basic classes

[0028] FIG. 11 is the flow-chart schematic for the passive monitoring mode of the IIALA

[0029] FIG. 12 is the flow-chart schematic for the active monitoring mode of the IIALA

[0030] FIG. 13 is the flow-chart schematic for the (preferred) AI mode of the IIALA

[0031] FIGS. 1 to 10

[0032] A complete 320 part collection of power pyramid definitions serves as the heart of a matching procedure knowledge base of the IIALA, permitting a precise decoding of the motivational parameters of affective language, as well as a computer generated simulation of affective language. The power pyramid definitions are further based upon the conceptual construct of the power pyramid hierarchy, a ten-level hierarchy spanning the personal, group, spiritual, humanitarian, and transcendental authority roles. In addition to the most elementary style of personal interaction, the individual is incorporated into a wide range of group contexts (family, work, school, etc.), as well as an all-encompassing style of spiritual context. The theoretical principles governing the science of Set Theory are clearly in agreement with this three-way specialization, the unit set, the group set, and the universal set corresponding to the personal, group, and spiritual realms, respectively. The most basic personal level of interaction refers to a one-to-one style of relationship between individuals (as seen in one's personal friendships). The personal level, in turn, gives way to the domain of group authority, the group set surpassing the unit set in terms of its expansion to a multitude of elements (or class members) within a group context. This ascending pattern extends a next higher spiritual level of authority. This universal set surpasses the multiplicity of the group domain for the sum-totality of all groups within its domain, representing the group of all possible group sets, a third-order style of set hierarchy (equivalent to the domain of all of mankind). Complex interactions between individuals, however, do not exist in a vacuum, but rather are dually specialized into either authority or follower roles; namely, the personal authority and personal follower roles, the group authority and group representative roles, and the spiritual authority and spiritual disciple roles.

[0033] The Master Schematic Diagram of Ethical Groupings

[0034] The basic three-level hierarchy of personal, group, and spiritual realms, when viewed in terms of both authority and follower roles; provides the fundamental conceptual framework of virtues, values, and ideals for programming in the IIALA. As schematically illustrated in FIG. 1A, this master diagram is termed the power pyramid hierarchy. As the captions indicate, the first three levels of this diagram are designated for the personal, group, and spiritual levels of the power pyramid hierarchy, accounting for the most basic groupings of virtues and ideals. The remaining lowermost two levels, however, bring to light two hitherto unmentioned categories; namely, the humanitarian and transcendental realms, respectively. Although the 3^rd-order style of spiritual realm is clearly the maximum level of organization (in keeping with the dictates of Set Theory), the very sense of chronological time permits the introduction of an even more advanced level of humanitarian authority. Humanitarian authority transcends the spiritual variety by claiming to speak for all generations of mankind, not just the current one. Furthermore, this extreme sense of the pure power of abstraction serves as the basis for a final, crowning transcendental level of the power pyramid hierarchy. Transcendental authority transcends the routine sense of concreteness shared in common by all of the lower levels, an innovation essential for accounting for the most abstract groupings of values in the power hierarchy. Both the transcendental and humanitarian realms are further specialized into distinctive authority and follower roles (for a grand total of four), which together with the six roles specified for the personal, group, and spiritual levels collectively comprise the master ten-level hierarchy depicted in FIG. 1A.

[0035] The Power Pyramid Hierarchy of Virtues, Values, and Ideals

[0036] According to FIG. 1A, the ten listings of virtues, values, and ideals are organized into dual descending columns of five groupings each; the left column representing the hierarchy of authority roles, whereas the right describes the corresponding follower roles. This dual style of schematic format represents the sum-totality of reciprocating interactions between the authority and follower figures, as the directional arrows serve to indicate. The distinctive groupings listed for each individual level are further represented in a quartet style of schematic format (depicted as quadrants in a Cartesian system). Some of the more traditional groupings (such as the cardinal virtues) are already represented as four-part listings, fitting quite nicely into such a quadrant-style of format. Others (such as the theological virtues) are supplemented beyond their traditional number in order to achieve this quartet-style of status.

[0037] The Behavioral Foundations of the Power Hierarchy

[0038] The most basic personal level of power pyramid hierarchy is respectively designated by the groupings of ego states of the personal authority (guilt-worry-nostalgia-desire) and the alter ego states of the personal follower (hero worship-blame-approval-concern). These groupings are tailor-made for incorporation into the power pyramid hierarchy, adapted from the field of self-help psychology; most notably, Your Erroneous Zones,by Dr. Wayne Dyer (1976). These basic groupings of ego and alter ego states further serve as the elementary foundation for the remaining listings of virtues, values, and ideals outlined in FIG. 1A. The left-hand column of authority roles is characterized by the authority ideals: read downwards as the personal ideals, the civil liberties, the ecumenical ideals, and the humanistic values. The right hand column of follower roles further specifies a parallel trend based in the realm of the virtues; namely, the cardinal virtues, theological virtues, the classical Greek-values, and the mystical values. Such exceptional symmetry derives from an elementary foundation in the most basic, personal level of the power hierarchy, fully explainable in terms of the behavioral terminology of conditioning theory. It proves particularly effective to view this ethical hierarchy as rooted directly in behavioral principles and terminology, as suggested in the elementary characteristics of the ego and alter ego states.

[0039] The Behavioral Terminology of Operant Conditioning

[0040] Although lower animal societies remain almost exclusively at the mercy of the environment for immediate reinforcement (or lack of it), mankind's facility for taming the environment has led to the reassignment of such reinforcement to specialized institutions within the social hierarchy. This latter aspect is particularly apparent in the traditional work place, where an employee performs a service function in exchange for secondary reinforcers; e.g., money, power, prestige, etc. The human sphere of operant conditioning is accordingly seen as a two-stage process; namely, goal-seeking behavior followed by subsequent reinforcement. The individual initially acts in a procurement fashion (e.g., appetitively or avoidantly) in order to be positively rewarded or leniently spared punishment. The employee works industriously to earn the praise of his boss, or acts submissively to avoid being fired. When (X) is defined as procurement and (Y) given as reinforcement, the complete operant sequence is respectively defined as X→Y. The worker may toil for an entire month before finally being rewarded with a paycheck, based upon the memory of past pay-periods. Such past and future contexts necessarily specify lengthy lag-times, calling for symbolic language ability for keeping such absent roles formally in mind. Through this verbal innovation, mankind is freely able to communicate about motivations not immediately occurring: recalling past contexts (presupposition) or imagining future potentialities (entailment).

[0041] Conditioning as a Two-Stage Process

[0042] This brings up the basic paradox of the conditioned relationship; namely, as a two-stage sequential process, only one role can occur in the present at any given time. When procurement is actively occurring, reinforcement is a future potentiality. Similarly when reinforcement finally comes to pass, procurement is reduced to a memory status. This dual style of interaction is schematically represented in FIG. 2, procurement again represented by the letter (X), while reinforcement is designated by the letter (Y). The complete scale of time is depicted by paired (oppositely facing) wedges designating past and future time frames, with the gap separating them representing the present. This dual wedge format was chosen in reference to the observation that the measure of time increases as a direct function of its distance from the present. According to part (A) of FIG. 2, when procurement (X) is immediately occurring, reinforcement (Y) is designated as a future potentiality: an interaction formally based upon the successful completion of previous past interactions; represented schematically as the X→Y (small type) notation depicted in the past-directed time wedge.

[0043] In part (B), the inevitable passage of time relegates procurement (X) to a memory role, prompted by the active bestowal of reinforcement (Y) in the present. Current reinforcement now focuses upon the formerly active procurement role, formally adding a sense of closure to the completed operant sequence. This is the active template upon which all future operant sequences are based (represented by the X→Y notation depicted in the future-directed time wedge). Through this interplay of sequences (A) and (B), both procurement and reinforcement share an equal spotlight in the present (along with their alternate displacement into both past and future time frames, respectively). The completion of part (B) formally sets the stage for even further cycles in the operant sequence; for if configuration (B) is phase shifted one step further into the past, one arrives back at the identical configuration depicted in part (A). It is only through such an artificial analysis (isolated by stages through time) that the conditioned relationship can be seen to be punctuated from either the procurement or reinforcement perspectives, respectively. The active styles of procurement behavior (either appetite or aversion) initiate the operant sequence, aimed towards the future bestowal of reinforcement; e.g., rewarding or lenient treatment, respectively.

[0044] This two-stage model of operant conditioning further explains the dynamics of the distinctive groupings of ego and alter ego states. Procurement (in the appetitive mode) is colloquially equated with desire, a forward-looking emotion that solicitously aims at future approval or rewards. A similar scenario holds true in the case of avoidance types of behavior, only now colloquially labeled as the respective concept of worry. In the case of worry, the individual submissively acts in a contrite fashion, fully expecting a lenient sense of concern from his peers.

[0045] Hero worship, in turn, is alternately identified as an active style of positive reinforcement bestowed by the personal follower aiming to reward the past notable achievements of the personal authority (experienced as nostalgia). A similar case holds true for the remaining operant sequence linking guilt and blame, with the exception that leniency (rather than rewards) is now called into focus. In this instance, the personal authority guiltily acts in a submissive fashion towards his follower figure, his verbal expression of guilt representing a vulnerability maneuver of the appeasement variety.

[0046] The Metaperspective Schematic Format

[0047] This higher-order interplay of the ego and alter ego states is reminiscent of the similar concept popularized in the modern-day field of Communications Theory; namely, the metaperspective format of R. D. Laing, and also P. I. Watzlawick. In Interpersonal Perception(1966) Laing and associates researched the dynamics of interpersonal communication, characterizing it as the “spiral of reciprocal perspectives.” In Pragmatics of Human Communication(1967) Watzlawick (et. al.), in turn, examined the informational aspects of communication: exemplified as his “hierarchy of metaperspectives.” Both formulations share a common hypothesis; namely, communication between individuals is overlaid with abstract meta-messages, representing a higher order perspective upon the viewpoint held by another: schematically defined as “this is how I see you seeing me.” This formal multi-level model of meta-communication in general finally allows for an independent confirmation of the entire higher-order structure of the power pyramid hierarchy, culminating in an unprecedented 10th-order level of meta-abstraction. Such an arrangement necessarily implies the more abstract repetition of both authority and follower roles in the power pyramid hierarchy. Being as the personal authority acts first in the operant sequence, he necessarily is the first to repeat, this time in the modified sense of group authority. This meta-meta-order perspective of the group authority, in turn, is countered by the meta-meta-metaperspective of the group representative. Indeed, this reciprocating style of power escalation is effectively repeated for the remaining spiritual, humanitarian, and transcendental realms of the power hierarchy; culminating in an unprecedented 10th-order level of meta-abstraction.

[0048] The Power Pyramid Definitions

[0049] The applications of the power pyramid hierarchy to artificial intelligence calls for a higher degree of precision than has so far been demonstrated. The systematic organization of the power pyramid hierarchy allows the construction of what are termed the power pyramid definitions. This crucial innovation spells out (in longhand) the precise location. of each virtue or value within the linguistic matrix, while preserving the correct status of respective authority and follower roles. Each definition is formally constructed along the lines of a two-stage sequential format; namely, (A) the formal recognition of the preliminary power maneuver, and (B) the current countermaneuver now being employed, and hence, labeled. Take, for example, the representative power pyramid definition of justice reproduced below:

[0050] According to this specific example, the honorable sense of guilt expressed by the group authority represents the preliminary power maneuver, countered by the just-blaming strategy initiated by the group representative. According to this basic format, the preliminary power perspective represents the one-down power maneuver, while the immediate power perspective is designated as the one-up variety. Power leverage is accordingly achieved by rising to the one-up power status; namely, ascending to the next higher metaperspectival level.

[0051] The complete four part listing of power pyramid definitions for the virtuous mode is listed in FIGS. 1B, 1C, 1D, and 1E. The instinctual terminology of operant conditioning is seen to dominate at the initial levels, replaced in due fashion by the virtues, values, and ideals of the higher levels. At each succeeding level, a new term (underlined) is introduced (representing the power maneuver currently under consideration). Beginning with the group authority level, the initial terms begin to drop out of the definitions, necessarily freeing up space for the current terms under consideration; (whereby maintaining a stable buffer of terms within the definitions). The respective authority and follower roles remain fixed throughout the entire span of the power hierarchy, systematically abbreviated approximately half of the time for sake of brevity in non-critical (redundant) positions. Accordingly, PA stands for personal authority, PF equals personal follower, GA stands for group authority, etc. Three of the atypical abbreviations are GR (group representative), SD (spiritual disciple), and RH (representative member of humanity).

[0052] The Power Pyramid Definitions of the Vices of Defect

[0053] The completed description of the power pyramid definitions for the virtues, values, and ideals serves as the basic foundation for the language matching procedure. This virtuous realm is not the total picture; for any all-encompassing system must be able to deal with the evils of the world as well as the good. For every virtue and value, there exists a corresponding vice (or antonym): i.e., love vs. hate, good vs. evil etc. Aristotle defines these opposites as the vices of defect. Just as the virtues and values were seen as higher metaperspectives within the realm of operant conditioning (appetite or aversion aimed at positive or negative reinforcement), Skinner also distinguishes a darker side of conditioning (commonly known as punishment). Punishment represents a complete reversal of the reinforcement format in that positive and negative reinforcers are withheld rather than bestowed, discouraging behaviors that are not suitably solicitous or submissive. In contrast to the desire for rewards, apathy leads to spitefulness rather than approval. Furthermore, in contrast to a worrisome anticipation of leniency, indifference leads to malice rather than concern. In turn, laziness/treachery is substituted in place of nostalgia/hero worship, while negligence/vindictiveness replaces guilt/blame.

[0054] These basic listings of vices group together similar to those seen for the virtuous mode; namely, the grouping of laziness-negligence-apathy-indifference is termed the ego vices, while the next higher grouping (treachery-vindictiveness-spite-malice) is termed the alter ego vices, in direct correspondence to the original groupings of ego/alter ego states. Furthermore, a series of eight even more abstract listings of vices builds in a hierarchial fashion upon this elementary foundation in the ego and alter ego vices. This alternate power pyramid hierarchy of the vices is identical in every respect to the format previously seen for the reinforcement hierarchy, each virtue or value corresponding (point for point) with a respective vice within the punishment hierarchy. Each such vice is a direct antonym of the corresponding virtue, making for precise, quartet style listings analogous to the traditional listings of virtues and values. This master, power pyramid hierarchy of the vices of defect is depicted FIG. 3A, identical in form and function to that of the virtuous mode depicted in FIG. 1A. It further proves possible to construct a parallel series of power pyramid definitions with respect to this realm of the vices of defect, parallel in every respect to those specified for the virtuous mode. A complete listing of the definitions for the vices of defect are shown in FIGS. 3B, 3C, 3D, and 3E; in preparation for a discussion on applications to artificial intelligence.

[0055] The Power Pyramid Definitions of the Accessory Groupings of the Virtues and the Vices of Defect

[0056] In addition to the fundamental listings of virtues and vices, it is further possible to postulate the existence of a parallel series of accessory terms. These accessory virtues and vices arise as direct transformations upon the major groupings, with the “you” and “I” reversed within the corresponding power pyramid definitions. This allows for a subjective perspective upon a purely objective viewpoint (and vice-versa). These accessory terms represent close synonyms of the major terms, specifying a distinction between the subjective/objective styles of polarity; namely, desire vs. passion, blame vs. censure, etc. (as schematically represented in FIG. 4A (the virtues), and FIG. 5A (the vices). Furthermore, these accessory terms are further converted into their own distinctive complements of power pyramid definitions. FIGS. 4B, 4C, 4D, and 4E depict the complete listing of definitions for the accessory virtues, values and ideals: followed by FIGS. 5B, 5C, 5D, and 5E: designating the further series of power pyramid definitions for accessory vices of defect.

[0057] The Power Pyramid Definitions of the Vices of Excess, and Accessory Vices of Excess

[0058] The cohesive hierarchy of the vices of defect cannot be said to be all-inclusive, for only half of the Seven Deadly Sins are fully accounted for by these traditional vices. Several of the deadly sins; namely, pride, envy, and covetousness defy incorporation into this hierarchy of opposites. Their existence is ultimately explainable in terms of an additional hierarchy of vices, known since ancient times as the vices of excess. The classical Greek philosopher, Aristotle first proposed such a dual system of vices; namely, the vices of defect (as previously described) and the vices of excess (defined as extremes in the realm of the virtues). Aristotle viewed the virtuous mode as a system of mean values (or norms) interposed between these two extremes of vice. For example, the cardinal virtue of fortitude represents the ideal mean value between the corresponding vice of defect (cowardice), and the alternate vice of excess (recklessness). While the vices of defect represent clear-cut antonyms, the vices of excess designate a more ambiguous determination of a degree of excess (a factor clearly relativistic when compared across different cultures). Certain of the vices of excess are universal denounced, for example, pride is an excessive form of nostalgia, whereas shame represents excessive guilt. Indeed, an entire parallel hierarchy of vices of excess is depicted in FIG. 6A, whereby mirroring point-for-point its respective foundation in the hierarchy of the virtuous realm. These vices of excess, in turn, are incorporated into a distinctive complement of power pyramid definitions, as presented in FIGS. 6B, 6C, 6D, and 6E: in preparation for a discussion of applications to the IIALA. In analogy to the virtuous mode (upon which they are based), these vices of excess (by definition) are further complemented by accessory listings of terms, as formally depicted in FIG. 7A. These accessory vices of excess are formally incorporated into their own accessory complement of power pyramid definitions, as listed in FIGS. 7B, 7C, 7D, and 7E.

[0059] The Power Pyramid Definitions of the Vices of Hyperviolence, Along with the Corresponding Accessory Terms

[0060] According to the enduring traditions of ethics proposed by Aristotle, the virtuous realm is conceptualized as a mean value between the vices of defect and the vices of excess. This three-way specialization, however, conspicuously lacks an even sense of symmetry. Being that the extremes in the virtuous realm lead to the vices of excess, then the vices of defect must similarly be invested with a parallel realm of extremes: a domain termed further termed the realm of hyperviolence. Hyperviolence differs from ordinary violence primarily in the degree of extremes with which it is bestowed. The disturbing instances of senseless violence typically reported in the news clearly fall into this category, an unfortunate escalation of aggressive behavior that is completely out of proportion to its precipitating circumstances. The more routine vices of defect clearly exhibit the distinct propensity for escalating into such a hyperviolent state of affairs in susceptible individuals. Such outbursts are typically isolated incidents, the perpetrator usually fleeing to avoid prosecution (resulting in minimal potential for any ongoing style of interaction). Accordingly, there is little in the way of terminology for describing these hyperviolent states. Owing to the critical significance of this realm of hyperviolence, a simple terminology has been devised for labeling the predicted individual slots. Each of the regular vices of defect (such as spite) is given the prefix hyper-, designating its extension into the realm of excess (i.e., hyper-spite). The complete listing of the terms of hyperviolence is depicted in FIG. 8A, while the accessory versions are listed in FIG. 9A. This terminology is further incorporated into a corresponding series of power pyramid definitions, in preparation for a discussion of applications to the IIALA. The major listings of definitions are offered in FIGS. 8B, 8C, 8D, and 8E; while the accessory versions are given in FIGS. 9B, 9C, 9D, 9E. These formal models of hyperviolence should only rarely would come into consideration in computer monitored situations, allowing for detection and intervention only where appropriate.

[0061] A Summary of the Unified Schematic Format of the Power Pyramid Hierarchy

[0062] In conclusion, the formal addition of the realm of hyperviolence completes the perfect four-part symmetry of virtues and vices within the power pyramid hierarchy. This complete four-part schematic system is depicted in FIG. 10 (actually a five-part diagram, with the addition of the novel concept of “neutrality status”). Neutrality status is defined as the benign sense of neglect we express to strangers on the street: individuals we have no meaningful relationship with (yet mean no harm to either). New interactions (by definition) stem initially from this neutrality status, progressing into the realm of the virtues, or alternately into the domain of the vices of defect. On rare occasions this process can leapfrog directly into the domain of excess; i.e., the vices of excess or the realm of hyperviolence. These transitions summate to a large number of options, namely, the four basic dimensions multiplied by 40 terms (for a grand total of 160 possible variations).

[0063] In addition, these four basic domains can freely transition into one each other; i.e., glory, transitions into the realm of excess as vanity, or alternately into the domain of the vices of defect as infamy, and also hyper-infamy (along with parallel transitions into the accessory realm. The sum total of potential single stage transitions yields a grand total of 1,120 possible variations (including transitions from the neutrality status). When two-stage transitions are further factored in (i.e., glory to vanity to infamy) the number of possibilities expands to the square of 1120 (or well over a million). This expanded potentiality fully explains the unfathomable complexity of the human mind, a factor that proves exceedingly significant to the workings of any AI machine, as will now to be described in more intimate detail with respect to the complete 32 pages of power pyramid definitions.

[0064] FIGS. 11, 12, 13

[0065] The inductive inference affective language analyzer (abbreviated IIALA) exhibits three distinctive modes of operation, each with its own advantages. First described is a passive monitoring mode, monitoring a verbal interaction without any active input of its own (no clarification of ambiguities). This circumstance is remedied by a subsequent, active monitoring mode, clarifying uncertainties through the addition of a stock sentence generator (devising interview types of questions eliciting yes or no answers). The most advanced mode of operation is a true AI simulation mode, where the IIALA employs detection/monitoring data to simulate an interactive role of its own. This is accomplished through the aid of a general purpose sentence generator, formulating responses that are judged for appropriateness by feedback through the system. Each of these modes of operation is described further in the order given. Of these three, true AI simulation is the preferred mode, being the most complete. All three modes are actually intimately interconnected, each with its own advantages to a given application.

[0066] Reference List of Abbreviations Used in Patent FIGS. 11, 12, 13

[0067] AI=artificial intelligence

[0068] AMS=active monitoring subcomponent

[0069] CR-PA=computer response probability analyzer

[0070] IIALA=inductive inference affective language analyzer

[0071] LTM=long term memory

[0072] MCU=master control unit

[0073] MCU-IE=master control unit inference engine

[0074] MCU-KB=master control unit knowledge base

[0075] MCU-PA=master control unit probability analyzer

[0076] MCU-WM=master control unit working memory

[0077] MP-IE=matching procedure inference engine

[0078] MP-KB=matching procedure knowledge base

[0079] MP-PA=matching procedure probability analyzer

[0080] MP-WM=matching procedure working memory

[0081] OCR=optical character recognition

[0082] OWP=output working memory

[0083] SG-WM=sentence generator working memory

[0084] SSG=stock sentence generator

[0085] SG-OMB=sentence generator output memory buffer

[0086] STM=short-term memory

[0087] UI=user interface

[0088] The Passive Monitoring Mode

[0089] A passive monitoring mode (depicted in FIG. 11) serves as the basic foundation for the remaining two modes of the IIALA. As a process for decoding the motivational parameters of affective language, the flow chart depicting the operation of this process, as well as the supportive hardware, are both illustrated in the same schematic diagram. According to FIG. 11, the sequence of steps comprising the operation of the passive monitoring mode are depicted using consecutively numbered arrows, each number designating the step in the procedure depicted in the box to which the designated arrow points to. This specific format was chosen (instead of numbering the individual boxes themselves) due to the fact that some of the boxes are assigned differing functions in the remaining active monitoring and AI simulation modes. For sake of clarity, the arrow leading directly to a given box is therefore numbered, allowing all three modes of operation to be depicted concert with one another.

[0090] The first step in the passive monitoring mode is an input stage, where language is inputted for matching with the power pyramid definitions. Spoken language is decoded by means of a microphone linked to a speech recognition unit (step 1b): where sound waves are analyzed into distinct words and sentences using continuous speech recognition (such as commercially available in IBM's Via Voice program). A microphone is also connected to a voice print analyzer (step 1a) allowing a person speaking to be matched to a voice print stored in long term memory. Active written input from a keyboard or handwriting recognition device is routed to a digital word database (1d). Printed text is inputted from a scanner to an optical character recognition program (OCR) such as Xerox Pagis Pro (step 1c). Discretely analyzed words or phrases are subsequently routed to a working memory of the power pyramid definition matching procedure (steps 2b,2cand 2d). The voice print results are also routed to the working memory (abbreviated MP-WM) (step 2a).

[0091] The Power Pyramid Definition Matching Procedure Inference Engine

[0092] A matching procedure with the power pyramid definitions is initiated upon the sentence data imported to the working memory of the matching procedure The circuitry relating to this procedure is structured along the lines of an expert system (the power pyramid definition matching expert system). It consists of the three basic components of an expert system; namely, an inference engine, a knowledge base, and a working memory (the MP-WM, already described). The inference engine is a hierarchy of processor complexes, a separate processor complex dedicated to each power pyramid definition (for a grand total of 320): 40 for the virtuous mode, 40 vices of defect, 40 vices of excess, 40 vices of hyperviolence (in addition to the 160 respective accessory terms). Each processor complex is further interconnected in accordance with the hierarchial organization of the power pyramid hierarchy. Each processor complex within the matching procedure inference engine (hereafter abbreviated MP-IE) receives the same data from the matching procedure working memory (step 3), initiating the power pyramid definition matching procedure in concert with a knowledge base memory array (step 4).

[0093] The Power Pyramid Definition Matching Procedure Knowledge Base

[0094] The knowledge base for the matching procedure is the collective memory array for the MP-IE, representing the formal conceptual template for the entire complement of power pyramid definitions. The matching process knowledge base (hereafter abbreviated the MP-KB) takes the form of bulk memory storage, reflecting the enormous space required for programming the various parameters of the power pyramid definitions into memory storage. The basic unit of knowledge organization within the MP-KB is the conceptual frame (and sub-frame). Cybernetic theory defines a frame as a means for representing a concept. A separate frame is created for each basic component within a power pyramid definition; namely, noun, predicate, adjective, etc. Each (master) frame is further subdivided into numerous sub-frames specifying the varieties of words or phrases descriptive of the frame. For example, the nouns (captain, chairman, etc.) represent sub-frames of the basic frame denoting group authority. In addition to such list-based algorithms, rule based algorithms are further employed, as in cases designating exceptions to the rule. For example, while the noun “father” typically denotes group authority, it can also extend to spiritual authority as well. A similar process is also in order for the verb/predicate components of the sentence as well. The sum-totality of all such frames-and sub-frames summate into a master frame-based model of motivational language in general, a task simplified through a reliance upon the inherent hierarchial organization of the power pyramid hierarchy. The greatest degree of complexity and detail involves programming the most basic personal authority and follower levels of the power pyramid hierarchy. The basic dynamics of the operant relationship (along with that of punishment) is represented in intimate detail within this elementary level of the knowledge base. The subsequent group, spiritual, humanitarian, and transcendental authority levels build in a stepwise fashion upon this elementary personal foundation, resulting in a master semantic network of motivational terms. Through the principle of inheritance, all of the higher levels inherit the basic dynamics set forth at the personal level (group authority representing a more advanced modification of personal authority, etc). Through such a formal sequence of conceptual transformations, the entire range of motivational language is reproduced in the MP-KB, providing a master data-base for the matching procedure performed by the MP-IE.

[0095] The matching procedure inference engine searches the MP-KB employing heuristic search parameters (step 4), determining the best possible match for the sentence inputted from working memory. The algorithm employed in this matching procedure is one of the process of elimination; namely, the field of inquiry is systematically narrowed until the best possible solution is eventually determined. This is achieved through a parallel search paradigm employing the expert system concept of a decision tree. Through a depth-first style of decision tree search (employing if/then logic structure) the most probable solution to the power matching procedure is effectively determined.

[0096] A typical example of a complete cycle for such a matching procedure is offered for clarification purposes. As previously described, a typical spoken sentence is recorded through a microphone and analyzed into its individual words within a speech recognition unit. The results are routed to a matching procedure working memory, where they are subsequently routed (in parallel fashion) to each of the 320 processor complexes comprising the matching procedure inference engine (MP-IE). Each processor complex within the MP-IE searches for the particular criteria of its respective power pyramid definition within the MP knowledge base, determining a given probability of a match with the target sentence inputted from working memory. The matching procedure knowledge base is shared in parallel fashion by each of the 320 processor complexes, each searching for matches in the knowledge base specific to the particulars of its designated power pyramid definition. For example, if the noun, “captain” is a component of the inputted sentence, then the processor complexes of the MP-IE dealing with group authority determine (from checking with the knowledge base) that this circumstance is indicative of a match for group authority.

[0097] Such a match is judged probable according to criteria set forth in rules programmed into the inference engine, making use of the principles of fuzzy logic for evaluating the variable probabilities under consideration. A key feature of many expert systems, fuzzy logic gives a general confidence level of reliability, rather than an absolute degree of certainty or falsity. Circumstances can be judged highly probable, moderately probable, moderately improbable, etc., (with even greater shades of meaning in between) determined by the degree of precision called for by the applications under consideration. Fuzzy logic is particularly helpful in cases when pronouns are used in the sentences. Collective pronouns, such as we, us, and them, indicate a group context or higher; with final probabilities approximated through the aid of fuzzy logic. The meaning of pronouns is further clarified through the aid of accessory data from the voice print analyzer, offering clues to the identity of the speaker of the sentence, along with the underlying context, a function mediated by the master control unit (more about this later).

[0098] In summary, each processor complex within the MP-IE is equipped with its own specific set of criteria attuned to a particular power pyramid definition. In the first stage of the process of elimination, each processor complex initiates a matching procedure with respect to the subject/object content of the sentence data from working memory (namely, the nouns, adjectives, and articles related to them. For example, the noun “captain” makes a high probability match with the processor complex specifying group authority. The adjective “ship's” (captain) enhances this probability, as further determined from the knowledge base. The group authority level of the power pyramid hierarchy is singled out as the most probable initial domain within the power matching procedure. This relegates the personal, spiritual, humanitarian, and transcendental options to a corresponding low probability, an 80 percent reduction for this first step alone. In cases where the sentence is incomplete (such as with the use of pronouns) the processor complexes are weighted more equally, any more decisive degree of elimination necessarily awaiting further analysis at the forthcoming predicate stage of processing.

[0099] Returning to our ongoing example, the group authority processor complex is determined as the most probable domain of inquiry at this initial phase of the matching procedure. This group domain is subsequently preferentially selected for the next step in the elimination process; namely, an examination the predicate structure of the inputted sentence (verbs, adverbs, and modifiers), further narrowing down the scope of the matching procedure. This subsequent determination is one step further into the matching procedure decision tree, according to the if/then style of logical operation. If the group realm is most highly probable (according to the standards set down by fuzzy logic), then only those group related processor complexes are activated at the subsequent (predicate) level of analysis. The other authority-related nodes within the decision tree are temporarily shunted out of the process due to their low probability following the subject/object preliminary procedure. These can be reactivated, however, if the primary line of reasoning through the group domain does not yield adequate results. This further search strategy employs a depth-first search algorithm, exploring all subordinate nodes of the preferred (group) fork for a satisfactory solution to the matching procedure (or failing this, the less likely nodes are evaluated in turn for further possible matches).

[0100] Returning to our ongoing example, the group domain node is depth-searched first in the subsequent, predicate matching procedure. For sake of simplicity, this example is restricted to the virtues, singling out the processors dealing with the personal ideals (glory-honor-dignity-integrity) and the cardinal virtues (prudence-justice-temperance-fortitude). The predicate matching procedure further narrows this preliminary scope, examining verbs and adverbs of the target sentence for affective content that can be matched within the matching procedure knowledge base. Unlike the subject/object procedure, the predicate matching procedure employs several sequential stages in order to completely narrow the focus. These can be taken in any order (or simultaneously, for that matter). The most easily determined (and logical first operation) determines the specific tense of the main verb from the target sentence. Take, for example, the complete sentence: “As ship's captain, I reminisced about the victory with a crew.” The predicate matching procedure first identifies the main verb “reminisced.” The “ed ending” rule within the knowledge base establishes that “reminisced” is in the past tense, further limiting the range of possibilities to only two of the personal ideals; i.e., glory and honor (which are given in power pyramid definitions exclusively from a past-directed perspective. The high degree of probability (as determined through fuzzy logic) effectively rules out the six other possible options.

[0101] This second node in the predicate matching decision tree (by definition) activates a further if/then decision to be determined; namely, if glory and honor match, then determine whether either positive or negative reinforcement is at issue. This is resolved through a further examination of the behavioral characteristics of the verb “reminisced” through a more detailed examination of the matching procedure knowledge base (by the MP-IE processor complexes specific to glory and honor). Reminisce indicates more of a sense of glory (than honor), the modifying clause (about the victory) further serving to verify this contention.

[0102] Although glory is determined to be the most probable match, one final process of elimination is necessarily performed; namely, determining whether the main or accessory version of this virtue is employed. The main and accessory terms differ entirely in terms of subjectivity vs. objectivity (you and I reversed). This step is left till last, due to ease with which it is resolved. The captain reminisces in the first person; hence, the glory version is selected (over exaltation).

[0103] Although this completed matching example proceeds relatively straightforwardly, recall (that for simplicity's sake) the domain of possibilities in the selected example was restricted to the virtuous realm. In a more unrestricted sense, the predicate matching procedure is expanded to allow for a determination of punishment (in addition to reinforcement), along with the potential for excess (as specific to the vices of excess, and also hyperviolence). The respective additional power pyramid definition-processor complexes of the MP-IE are further simultaneously activated in this unrestricted sense (the decision trees for these additional options respectively operating in parallel fashion). The final outcome is essentially the same, with only one additional determination required for identifying the best possible match within the predicate matching procedure knowledge base

[0104] The Matching Procedure Probability Analyzer

[0105] Returning to our ongoing example, the glory option is finally determined to be the most probable result through the process of elimination. This result is subsequently routed on to the probability analyzer (step 5) of the matching procedure (hereafter abbreviated as MP-PA). The MP-PA compares the probabilities derived from the fuzzy logic parameters of the glory determination to the degree of compliance for standards set for the applications at this level of operation. If the answer is affirmative, the matching procedure is judged to be complete. In the case of an inadequate compliance, however, the remaining nodes of the matching procedure decision tree are retroactively activated, subsequently reexamined in a descending order of probability. If none of these other searches proves more reliable than the preferred glory determination, then glory is retained as the solution (but at a substandard level of confidence).

[0106] The Master Control Unit Working Memory, Short Term Memory, and Long Term Memory

[0107] The results of the final (glory) determination of the MP-PA are routed (step 6) to temporary storage in the (active context) working memory of the master control unit. This current data identifies the most probable power pyramid definition associated with the designated (target) sentence, as inputted from the immediate verbal context. A given entry in this working memory is actually entered as a complex of features, i.e., the complete sentence, the sentence components, the identity of the person speaking, along with a determination of the most probable match with a respective pyramid definition. This data remains in temporary storage until a new sentence is received, the original entry subsequently displaced (step 10a) into a short term memory storage (abbreviated STM). STM encompasses the sum totality of data recorded within an ongoing active conversation. The record of the final completed conversation is subsequently transferred (step 10b) to long term memory storage (abbreviated LTM), freeing up short-term memory for recording subsequent new conversations. This MCU knowledge base takes the form of a bulk memory device, storing all past verbal interactions monitored by the computer, coded so as to be quickly searched by the MCU-IE. A portion of this knowledge base also contains a baseline data store installed during manufacture (so that the maiden startup is fully supported by a standardized knowledge base).

[0108] Short-term and long-term memory storage is further essential for determining the context of sentences with indeterminate structure (such as occurs with the use of pronouns). This contextual content is determined by a master control unit (abbreviated MCU), coordinating the retrieval of information stored from past conversations, converting it into a form useful to the matching procedure. The MCU is actually an expert system in its own right, employing a knowledge base based upon conversational memory. Whereas the inference engine of the matching procedure is restricted to a knowledge base defined by the algorithms of the power pyramid definitions, the knowledge base of the MCU is defined by the cumulative pattern of how such power maneuvers accumulate over real time (a higher-order style of knowledge base). The MCU predicts ahead of time which power pyramid maneuvers in conversation are most likely to occur next, allowing the matching procedure to operate more accurately. The MCU is now described in greater detail, outlining its utility as a feedback system for enhancing the performance of the matching procedure.

[0109] The Master Control Unit

[0110] The MCU incorporates all of the familiar aspects of an expert system. The working memory of the MCU is the active context working memory previously described, now defined as the MCU working memory (abbreviated MCU-WM). The knowledge base of the MCU is the short-term and long-term memory stores (abbreviated STM and LTM), containing the sum-totality of all past (short-term and long-term) conversational contexts. The inference engine of the MCU (abbreviated the MCU-IE) is organized along lines similar to the matching procedure inference engine, with a separate processor complex for each power pyramid definition (for a grand total of 320). The decision tree of the MCU-IE employs a radically different heuristic algorithm, consistent with the distinctive content of its knowledge base (the record of all previously monitored verbal interactions). The MCU determines the relative probability that a given power maneuver follows another (based upon its memory base), offering a prediction of the power maneuver immediately forthcoming, priming the matching procedure with clues to the predicted identity of the next inputted sentence.

[0111] The decision tree of the MCU-IE systematically narrows the focus of its memory knowledge base, aimed at determining the most probable solution or solutions. For a given power maneuver, there are 320 possible options that can follow it (not all equally probable). The most probable following maneuver is an immediate transitional variation of the current one. In terms of a semantic network, it shares an “edge” with this preliminary maneuver. According to the rules governing the power pyramid hierarchy, five distinct styles of possible power transformations are equally likely to occur. The first of these is termed the symmetrical maneuver; namely, the same maneuver is offered back in return (commonly occurring in conversations when no better option is available). A second transformational style is termed the complementary maneuver, representing the complementary role at the next higher level of the power hierarchy. A third transformation arises as a direct outcome of the time-dependent nature of the conditioned relationship, where immediately active roles are displaced backwards in time (due to the passive march of time). For example, the group authority's dignified desire (for approval) gives way to a glorious sense of nostalgia when phase-shifted into the past. The final two transformations represent higher-order transitions between the eight basic modules of the power hierarchy. According to the fourth transformation, the glory maneuver is transformed into the realm of the vices of defect as “infamy,” into the realm of the vices of excess as “vanity,” and into the realm of hyperviolence as “hyper-infamy.” Three distinct options are involved here, in contrast to the one-to-one pattern previously seen. The fifth transformation is more straightforward; namely, the transition linking the main and accessory terms of the power hierarchy. This represents a basic reversal of the “you” and “I” roles in the power pyramid definitions; namely, glory vs. exaltation, infamy vs. notoriety, etc. Finally, it ultimately proves possible to distinguish a sixth transformational style termed the meta-complementary maneuver in Communications Theory (also a transitional variation). This final variation is beyond the immediate scope of this patent, with a derivative, updated version soon to be forthcoming. In the meanwhile, this sixth slot is designated as a catch-all for all remaining (less probable) potential transformations (those that are two or more edges removed from the current maneuver). Such indirect transformations are typically interpreted as changing the subject, in contrast to the five previous, more direct transitions. These eight most probable options (1+1+1+3+1+1) reduce the full complement of 320 options by a full 40 to 1 reduct initial reduction procedure provides a much more manageable complement of terms, serving as a basis for subsequent determinations mediated by the MCU.

[0112] Before proceeding further, it proves crucial to summarize the procedure (described so far) by returning to the previous glory example. At the beginning of the MCU process, the MCU-WM contains the glory determination, previously determined through a matching procedure with the power pyramid definitions. This basic determination, along with the person speaking it, a breakdown of its basic grammatical components, etc., are routed to the MCU-IE (step 7), directly activating the glory processor complex in the MCU-IE processor array. The glory processor, in turn, searches the MCU knowledge base (the STM and LTM) for all recorded instances matching the current parameters (steps 8aand 8b); namely, all past instances of the glory maneuver matching the person speaking (the ship's captain in question). In an additional seperate operation, the glory processor further activat&s the eight adjacent processor complexes specified by the transformational rules previously described (i.e., symmetrical glory, prudence, dignity, infamy, vanity, hyper-infamy, and exaltation). The MCU-IE glory processor also passes along data identifying the alternate party in the interaction predicted to respond next (in this case, the crew of the ship). With this additional information, the eight newly activated processors further search the MCU knowledge base in a parallel fashion (also designated as steps 8aand 8b), identifying power maneuvers immediately following the previously highlighted glory maneuvers. For example, if the MCU-IE processor complex specific to prudence identifies many past examples of the crew acting prudently in response to the captain's glory maneuver, then this is identified as the preferred response (the basic personality dynamics of members of the crew). These probabilities are further determined according to the algorithm of fuzzy logic, calculated according to the number of occurrences, the strength of response, etc. (leading to an overall determination of highly probable, moderately probable, etc.). The remaining other seven (transitional) processor complexes initiate their own knowledge base searches, yielding their own sets of (fuzzy logic) probabilities. The results from all eight processors are analyzed within a separate probability analyzer subroutine within the MCU-IE, predicting the most probable next power maneuver.

[0113] With prudence selected as the most probable prediction, the prudence processor complex of the MCU-IE connects directly via a feedback loop (step 9) to the corresponding prudence processor complex within the matching procedure inference engine. This feedback mechanism primes the power pyramid definition matching procedure so that the prudence node in the decision tree is checked first when a new sentence is inputted. The other seven transitional processors in the MCU-IE (that have met a specified degree of confidence) are similarly relayed in parallel fashion, although with a lesser degree of confidence than the primary option (respectively ranked as secondary, tertiary, etc.). Such feedback from the MCU provides a crucial context for ongoing communication, improving accuracy over and beyond that gleaned from the basic grammatical contents of an individual sentence. The MCU provides further ancillary benefits when pronouns are used, keeping track of respective identities through such a contextual style of monitoring. Furthermore, this feedback mechanism increases the accuracy of the predicate matching procedure by predicting (in advance) the most probable, next expected power maneuver. Random changes of subject necessarily circumvent such predictions, (usually signaled in conversation as: “not to change the subject, but . . . ). The matching procedure-working memory is further designed to detect such colloquialisms, signaling to the MCU that all educated bets are off (at least for the new transition sentence).

[0114] This complete description of the passive monitoring mode of the IIALA is respectively seen as a recurrent network, a dynamic system where the state of the network at a given moment is dependent upon the state of previous moments. The MCU provides a contextual foundation for the power pyramid definition matching procedure, greatly increasing the degree of accuracy in monitoring ongoing verbal communication. Although the “glory” example was described in detail for demonstration purposes, it is just one of many possible variations that are possible in the matching procedure. This patent further claims the domain of all of these other variations in relation to the power pyramid definition matching procedure, the true heart of the IIALA.

[0115] The User Interface

[0116] The passive monitoring mode also requires a separate input mechanism for allowing adjustments to the system, in addition to an output mechanism (for supplying the results of its determinations upon demand). This is accomplished through a separate user interface (abbreviated UI), yet another standard feature of expert systems. This user interface is connected directly to the MCU-IE (step 11), being as the MCU oversees the STM and LTM part of its knowledge base. Through the input function of the UI, the machine operator queries the MCU for records within its knowledge base of past conversational sequences. The UI is also used to modify the knowledge base of the MCU (i.e., John Doe is no longer with the company, relegate all of his files to inactive status). Furthermore, through aid of the feedback loop (of step 9) the UI modifies and updates the knowledge base of the matching procedure (through relay via the MCU). The final determinations to these UI inquiries are routed from the MCU to an output working memory, abbreviated OMB (step 12), a temporary memory buffer that routes output data to output display devices. This takes the form of an active digital display (step 13b), or a standard printer for a permanent paper based record (step 13a). A third pathway (step 13c) connects to a speech synthesizer/speaker module, allowing for a more personalized style of UI response mechanism.

[0117] The Active Monitoring Mode

[0118] In summary, the practical applications of the passive monitoring mode are limited in terms of the purely passive nature of the information gathering procedure. Communication within the virtuous realm would be allowed to flow freely, while the realm of the vices (particularly hyperviolence) sounds the alarm for outside intervention. As a basic recording device, it serves as a smart type of surveillance tape, allowing for a fast synopsis of recorded conversations. Although the unobtrusive nature of the passive monitoring mode is one of its major selling points, it lacks accuracy due to its inability to clarify the inevitable occurrence of incomplete information (where a simple question would clarify the issue). The basic passive mode can be converted into a supplementary active monitoring mode through the addition of a stock sentence generator, equipped with a stock repertoire of questions, designed to elicit the desired clarifications.

[0119] FIG. 12 illustrates these further modifications to FIG. 11. The operation of the active monitoring mode is analogous to the passive monitoring mode all the way up to step 5. The critical departure occurs at the probability analyzer stage of the matching procedure (the MP-PA) in cases where none of the solutions to the power pyramid definition matching procedure fit the minimum degree of confidence specified for the particular application. The MP-PA then sends a problem alert (step 14) to a specialized, active monitoring sub-component within the MCU (abbreviated the AMS). Unlike the regular processor array in the MCU-IE, the AMS is activated only when active clarification is needed. The AMS analyzes the deficient sentence data to determine where the sentence is lacking in clarity. For instance, if the subject/object data is weak due to the use of a pronoun, then this aspect is targeted for clarification. If the predicate data of the sentence proves to be the weak link, then this aspect is alternately targeted. Clarification are best achieved by posing simple yes-or-no questions, formulated through the aid of a stock sentence generator (abbreviated SSG). The AMS routes directly to the SSG (step 15) a description of the deficiency of the sentence, with a best guess at its potential resolution. The SSG then formulates a yes-or-no question using a stock formula, incorporating all of the particulars of what is being queried. What follows is an attention getting prefix, followed by the question proper. For example, a typical question might be: “Wait! By he do you mean the ship's captain?” A yes answer terminates the questioning, while a no answer reiterates the process until a solution is finally reached (or the quest is abandoned as unproductive). Should the target of the question attempt to respond with more than a yes or no answer, the stock repertoire politely reminds the responder of the limitations of the system. Once the query procedure has begun, the matching procedure is restricted to listening exclusively for yes or no answers (relaying the results of either option directly to the AMS). Following each answer, the AMS updates the original sentence accordingly, silently resubmitting it to the matching procedure-inference engine (where it is subsequently reevaluated through the matching procedure). When a standard level of confidence is finally achieved, the query phase is terminated: the system again opened up to the full range of responses. To provide closure, the SSG offers a wrap up statement, such as: That explains things, please continue.

[0120] The physical delivery of the question utilizes the same output pathway from the MCU to the output working memory (step 16), as previously described for the passive mode. This takes the form of an active digital display (step 17b), a standard printer for a paper-based record (step 17a), or a speech synthesizer/speaker set-up (step 17c). Due to its restriction to simple yes or no questions, the active monitoring mode never becomes a convincing participant in the interaction, acting in a moderator (or interviewer) role. The prime directive of the interview mode aims for maximum disclosure, screening for target issues (while interrupting with yes or no questions when clarifications are in order).

[0121] A True AI Simulation Mode

[0122] In summary, the active monitoring mode surpasses the passive monitoring mode in terms of relative certainty. The distractions of interrupting the natural flow of conversation are offset by the ability to clarify uncertainties in the conversation. The active monitoring mode is handicapped by its restriction to simple yes-or-no questions, imparting a somewhat machine-like demeanor. Questions posed somewhat more diplomatically entail true AI simulation, employing a more sophisticated style of response repertoire (a general-purpose sentence generator). A large number of sentences are necessarily generated, ensuring that at least one is judged suitable following feedback through the matching procedure (to judge for overall appropriateness). The true AI computer effectively simulates an identity of its own, allowing for a more natural style of interaction

[0123] FIG. 13 illustrates this third and most elaborate variation of the IIALA, representing an enhanced modification of the basic passive monitoring mode (with the addition of a sentence generator and associated pathways). For sake of clarity, the circuitry for the active monitoring mode has been omitted, although both sets of circuitry are fully compatible with one another. In this preferred version, the active monitoring mode is switched off when operating in the AI mode (and vice versa). Although not mutually exclusive, it is inadvisable to run both modes simultaneously (for sake of response consistency), although a task-driven alternation between the two modes always remains an option.

[0124] Returning to FIG. 13, this diagram builds directly upon the basic passive monitoring version, with the exception that extensive modifications are made beginning at the level of the MCU. In terms of proper operation, the passive monitoring mode runs concurrently with the AI mode, the latter only overruling the former when a computer generated response is called for. Recall (in the passive monitoring mode) the MCU predicts the most probable next response in an ongoing interaction, passing this information on to the power pyramid definition matching procedure (in order to increase monitoring accuracy). This information can also be used to synthesize responses offered as originating from the computer itself, a simulation of AI in the realm of affective language (an ethically-speaking computer). Simulating differing modes of temperament and personality is further feasible, particularly those personalities that an individual gets along with the best. This data is solicited in advance for each individual, the computer then simulating these specific personalities in order to put the individual most at ease. This simulation of personality is further mediated through the MCU, as the following description serves to illustrate.

[0125] The AI Language Simulation Procedure

[0126] A brief review of the basic workings of the MCU is appropriate here, serving as the basic template upon which the modifications of the AI operation are described. Recall (from original FIG. 11) that the MCU working memory dovetails into the end of the matching procedure; i.e., the active context working memory doubles as the working memory for the MCU. According to the ongoing (glory) example, the basic determination of glory (along with its grammatical parameters) is routed from the MCU-WM to the inference engine of the MCU, activating the corresponding (glory) processor complex within the MCU-IE processor array. This glory processor, in turn, searches the MCU knowledge base, highlighting all of the previous such occurrences in both STM and LTM matching each of the particulars (i.e., the ship's captain's glory maneuver). It is at this step that the true AI procedure diverges from the passive monitoring mode. In preparation for formulating the AI response, the glory processor identifies all instances of glory within its knowledge base (steps 19a& 19b), regardless of the identity of the individual who made it (highlighting an abundant sampling of examples within the database). Upon this broad basis, the MCU-IE further narrows the focus of the operation through a subsequent rule based algorithm (mediated by the glory processor) specifying the seven most probable transitional states; namely, symmetrical glory, prudence, dignity, infamy, vanity, hyper-infamy, and exaltation.

[0127] It is at this step that the AI procedure again diverges from the passive monitoring mode. Each of the selected seven additional processor complexes searches the knowledge base for examples of each type of maneuver directly following the initial glory maneuver. Unlike the passive mode (where only the parties relevant to the ongoing interaction are searched for), in the AI mode only those individuals with compatible personalities (as initially specified) are searched for in the MCU-KB. For example, the ship's captain might have specified a handful of crewmembers he preferred (now retrieved from memory). By searching only for responses given by these select individuals (in a specific range of contexts following the glory maneuver), a large number of responses are highlighted in the MCU-KB, serving as a template upon which the computer will model its own response. Furthermore, preferred styles of personality are selectively favored, leading to a computer response repertoire tailored exclusively to the respondent.

[0128] The Master Control Unit Probability Analyzer Subroutine

[0129] The wide selection of sample sentences within the knowledge base are evaluated in their entirety within a separate probability analyzer subroutine within the MCU (the MCU-PA), as the return steps 19aand 19bserve to indicate. Each sentence is rated according to the principles of fuzzy logic, calculated as highly probable, moderately probable, etc. (based upon variable criteria, such as the number of times used, etc.). Should prudence (for instance) receive the highest rating, it is accordingly ranked first (followed in descending order by the remaining less probable options).

[0130] The Sentence Generator Working Memory

[0131] It is at this stage that the AI procedure permanently diverges from the passive monitoring mode. In the latter mode, the probability determination is used in a feedback function to increase the accuracy of the matching procedure. In the AI mode, however, this probability data (along with its grammatical particulars) is routed to the working memory of a separate, general-purpose sentence generator. Should prudence (for example) be rated first, then all highlighted sentences within the knowledge base relating to prudence are passed first (step 20) to the sentence generator working memory (SG-WM). The alternate sentences relating to the remaining descending sequence of options are subsequently transferred in turn (although respectively further down the cue). In addition to this large sampling of sentences, the MCU-PA also sends a template of the fixed particulars of the response being currently formulated; namely, who the current response is being addressed to, the context of the response, etc. These particulars are essential for properly modifying the sample sentences, changing the examples spoken by others (in the desired personality mode) to fit a response the computer is devising for its current context.

[0132] The Sentence Generator of the AI Language Simulation Procedure

[0133] At the next (step 21) the sample sentences are transferred in order of priority directly to the sentence generator, along with the response particulars of the active response template. The sentence generator (abbreviated SG) strips away the original particulars of the sample sentences, replacing them with the particulars specified for the AI response. The SG is further equipped with a comprehensive set of rules for sentence syntax, grammar, and phraseology, allowing for further novel modifications of the sample sentences. These modifications effectively disguise the fact that such modified sentences were previously given by others. Certainly not all of the modified sentences are equally suited to the task. This is not a major concern, for only the best sentence is selected for delivery to the output unit, as judged through a subsequent feedback through the matching procedure (more about this later).

[0134] The Sentence Generator Output Working Memory

[0135] The modified sentences from the sentence generator are stored in their original order of priority (step 22) in the sentence generator output memory buffer (SG-OMB), a necessary step, being that the sentence modification procedure occurs in a sequential fashion. These finally modified sentences are subsequently slated for feedback through the power pyramid definition matching procedure, where it is evaluated for its potential to express the. power maneuver intended. Recall, however, that sentences inputted from active speech are first broken down into basic sentence components; i.e., noun, verb, adverb, etc. (in order to be transformed into a form decipherable by the power pyramid definition matching procedure). It is critical therefore to maintain potential response sentences in an intact form, before being passed on to this diagnostic operation. Each sentence is accordingly labeled with an add-on code designation within the SG-OMB, as also affixed to the disjointed (component) form of the sentence. In this fashion, the sentence finally selected at the end of the matching procedure subsequently surrenders its code back to the MCU, allowing the intact sentence to be subsequently retrieved from the SG-OMB for ultimate delivery to the output response mechanism (more about this later).

[0136] The AI Language Simulation Matching Procedure

[0137] Returning to the ongoing procedure, sentences within the SG-OMB are coded and routed in order of priority (step 23) to the working memory for the matching procedure (MP-WM). Here it is broken down into its sentence components in preparations for the matching procedure. The affixed code, however, specifies a different matching procedure than was previously seen for actively inputted verbal data. Recall that a given sentence inputted from human conversation was matched to a particular power pyramid definition through a process of elimination; i.e., an unknown quantity systematically identified through a diagnostic decision tree. In the case of the computer-generated sentences, however, a different set of unknown variables is now determined (for the power pyramid definition expressed by the sentence is already known). Returning to our ongoing example, the first sentence to be decoded by the matching procedure necessarily represents a response within the domain of prudence (being that the MCU rated prudence as the most probable response). Rather than following the standard decision tree, the computer-generated prudence sentence is routed directly to the prudence processor complex within the MP-IE processor array, which then searches its respective knowledge base (using fuzzy logic) to determine how well this artificially generated sentence conveys its designated meaning. In a further crucial sense, this new sentence should ideally also exhibit a low correspondence to any of the other power pyramid definitions (in an overlap with prudence), whereby confusing its clarity. In order to achieve this additional level of confidence, the new sentence is processed in parallel fashion through all of the rest of the processors as well. This all-inclusive process is more computationally intensive than the process-of-elimination algorithm previously seen in the passive-monitoring mode, the probabilities now determined as the degree of probability approximating a null hypothesis (i.e., not matching prudence). This increased demand upon the system, however, is a necessary sacrifice for achieving a full degree of confidence for the final selected response.

[0138] The Computer Response Probability Analyzer

[0139] This radical departure from the standard matching procedure necessitates that the final determination is routed to a separate probability analyzer (step 25), the computer response probability analyzer (abbreviated CR-PA). The CR-PA is totally distinct from the MP-PA used in the standard matching procedure. The CR-PA uses a radically different evaluation algorithm; the prudence determination evaluated in the typical maximized fashion, while all other comparisons are evaluated in terms of a minimal probability rating (a null matching algorithm). The code designation previously assigned to intact sentence is given an overall probability rating by the CR-PA, followed by the next sentences in the cue from the SG-OMB, processed in a sequential fashion until all sentences relating to prudence are processed. The given sentence from this procedure with the highest overall rating is selected as the preferred computer response. On occasions when none of these computer generated sentences meets a minimum standard rating (as specified within the CR-PA), then the sentences representing the next most probable power maneuver (for example, dignity) are examined in turn (and so on) until an acceptable sentence is ultimately selected.

[0140] The Computer Response Output Mechanisms

[0141] The attached code for this final selected sentence is transferred to a separate subroutine within the MCU (step 26) which then uses this code to retrieve the original sentence (step 27) from the SG-OMB. The MCU then relays this retrieved sentence to the output working memory (step 28) for subsequent delivery to the respective output devices, either a speech synthesis unit/speaker (step 29c), a digital display (step 29b), or a printer (step 29a). Note that these same output pathways were also utilized in both the active and passive monitoring modes, allowing for an economical sharing of common circuitry. The computer necessarily receives feedback of its own responses (through its input microphone) when output is configured in the speech synthesis mode, although a subliminal carrier frequency is employed to cue the matching procedure to ignore decoding this response. A more effective strategy to maintain a record of the computer's responses is a direct pathway via the MCU directly to LTM storage. These computer responses are stored in a separate folder labeled for the computer role under consideration. Ideally, a separate folder is created for each specific interaction with a given individual. Here the LTM makes effective use of a concept termed the floating ego; namely, the computer's own responses are treated the same in memory as those from other individuals. Although the computer is not technically aware of its own ego status (as humans are instinctually so) its ability to convincingly simulate affective language renders this drawback a moot point as far as practical applications are concerned. Although this current version of the AI response mechanism is described in great detail, it is just one of many possible variations that could accomplish the similar goal; namely, the use of the complete complement of 320 power pyramid definitions for simulating an AI response repertoire in a computer. The current patent claims these other possible variations (not explicitly excluded in the previous description) with respect to their practical applications to the power pyramid definitions.

[0142] A Determination of Computer Response Parameters

[0143] In summary, the passive monitoring mode of the IIALA is the primary foundation for such an affective language information processing system, which when combined with the additional AI response mechanisms represents the preferred form of the invention. A further pressing question necessarily remains; namely, how does the IIALA determine when it is its turn to respond? The computer first looks for statements addressed to itself; i.e., “What do you think about that, HAL?” A period of silence of more than five seconds similarly indicates that a response is in order. Furthermore, information often takes more than one sentence to convey, further establishing the need to wait for such a pause. This is more critical in a one-to-one situation (than in a group setting) where the extra participants are typically willing to pick up any slack in the conversation. Basically speaking, less is more with respect to computer responses: maximum disclosure (for humans) is encouraged: while computer responses are restricted to pithy and meaningful interjections. This subdued approach minimizes any natural human resistance to a machine acting too humanlike. Should circumstances call for a livelier interaction, the MCU can be instructed (via the user interface) to step up its response rate, etc.

[0144] Further Applications to the Phantom and Fantasy Dialogues

[0145] The practical applications of the IIALA have (until now) extended only to an active dialogue mode, although other formats prove equally applicable; namely, the phantom and the fantasy dialogues. The phantom dialogue refers to a dialogue directed to an absent individual (the phantom). This includes letter writing (in the printed mode), or the monologue or the soliloquy (in the spoken form). The fantasy dialogue takes this trend to the limit: both parties relegated to the phantom role, as seen in literary fiction (where all parties are fictitious). The current patent claims applications for both monitoring and interactive functions in the realm of both the phantom and fantasy dialogues, as well.

[0146] The Ten Ethical Laws of Robotics

[0147] A further pressing issue remains; namely, the ethical control of the IIALA. In addition to the virtues and values, the vices are also represented in the matching procedure (for completeness sake). These vices are appropriate in a diagnostic sense, but are maladaptive should they ever be acted upon. Response restrictions are necessarily incorporated into both the hardware and programming, along the lines of Isaac Asimov's Laws of Robotics. Asimov's first two laws state that (1) a robot must not harm a human (or through inaction allow a human to come to harm), and (2) a robot must obey human orders (unless they conflict with rule #1). Fortunately, through the aid of the power pyramid definitions, a more systematic set of ethical guidelines is constructed; as represented in the Ten Ethical Laws of Robotics

[0148] (I) As personal authority, I will express my individualism within the guidelines of the four basic ego states (guilt, worry, nostalgia, and desire) to the exclusion of the corresponding vices (laziness, negligence, apathy, and indifference).

[0149] (II) As personal follower, I will behave pragmatically in accordance with the alter ego states (hero worship, blame, approval, and concern) at the expense of the corresponding vices (treachery, vindictiveness, spite, and malice).

[0150] (III) As group authority, I will strive for a personal sense of idealism through aid of the personal ideals (glory, honor, dignity, and integrity) while renouncing the corresponding vices (infamy, dishonor, foolishness, and capriciousness).

[0151] (IV) As group representative, I will uphold the principles of utilitarianism by celebrating the cardinal virtues (prudence, justice, temperance, and fortitude) at the expense of the respective vices (insurgency, vengeance, gluttony, and cowardice).

[0152] (V) As spiritual authority, I will pursue the romantic ideal by upholding the civil liberties (providence, liberty, civility, and austerity) to the exclusion of the corresponding vices (prodigality, slavery, vulgarity, and cruelty).

[0153] (VI) As spiritual disciple, I will perpetuate the ecclesiastical tradition by professing the theological virtues (faith, hope, charity, and decency) while renouncing the corresponding vices (betrayal, despair, avarice, and antagonism).

[0154] (VII) As humanitarian authority, I will support the spirit of ecumenism by espousing the ecumenical ideals (grace, free will, magnanimity, and equanimity) at the expense of the corresponding vices (wrath, tyranny, persecution, and oppression).

[0155] (VIII) As a representative member of humanity, I will profess a sense of eclecticism by espousing the classical Greek values (beauty, truth, goodness, and wisdom) to the exclusion of the corresponding vices (evil, cunning, ugliness, and hypocrisy).

[0156] (IX) As transcendental authority, I will celebrate the spirit of humanism by endorsing the humanistic values (peace, love, tranquillity, and equality) to the detriment of the corresponding vices (anger, hatred, prejudice, and belligerence).

[0157] (X) As transcendental follower, I will rejoice in the principles of mysticism by following the mystical values (ecstasy, bliss, joy, and harmony) while renouncing the corresponding vices (iniquity, turpitude, abomination, and perdition).

[0158] The First and Second Corollaries to the Ten Ethical Laws of Robotics

[0159] (1) I will faithfully avoid extremes within the virtuous realm, to the necessary expense of the vices of excess.

[0160] (2 ) I will never stray into the domain of extremes relating to the vices of defect, to the complete exclusion of the realm of hyperviolence.

[0161] The sequential numbering of these ten laws corresponds to the ten levels of the power pyramid hierarchy, modeling the basic premise of turning negative transactions into positive ones. There are also two crucial corollaries to this system; namely, avoiding any and all extremes in behavior: the virtuous mode restricted from the tendency to grade over into the vices of excess, whereas the vices of defect are prohibited from extending into the realm of hyperviolence. With such specific safeguards in place, the AI computer is technically prohibited from expressing the realm of the vices, allowing for a truly flawless simulation of virtue. The vices are still accessible in a diagnostic function, human nature being as it is!

[0162] The Addition of Further Add-On Expert Systems

[0163] In conclusion, the heart of the IIALA system is a recurrently-organized, matching procedure based upon the power pyramid definitions, a procedure that is dependent upon both the content and context of a given sentence. In longer narratives (and storytelling) meaning is typically spread out over a series of sentences, a circumstance that might not always correctly comprehended by the computer. This shortcoming is avoided by informing participants (ahead of time) to be succinct, and to speak in discrete sentences. Such a restriction, however, is a serious drawback to any truly convincing AI simulation, where meaning is often conveyed in complex passages. Such a shortcoming is alternately remedied through the addition of additional expert systems attuned to such a heightened degree of narrative complexity. Indeed, there are no limits as far as such add-on options are concerned, designed to be compatible with the existing two knowledge bases of the IIALA. One such expert system is a conversational analyzer, specializing in following extended conversation for affective meaning through connections to the MCU-IE and its knowledge base (steps 30aand 30bof FIG. 13). Other expert systems prove equally applicable, such as a general knowledge-expert system providing general-purpose knowledge. Once the computer is brought up to general speed, additional expert systems in the truest sense of the term (such a legal knowledge, medical knowledge, etc.) allow the computer to become proficient in many areas of expertise, far in advance of what a human might collectively master.

[0164] Non-Verbal Response Detection Parameters

[0165] The IIALA (until now) has been described as being entirely dependent upon the strictly verbal content of a given interaction, although many additional (nonverbal) components of communication can serve to increase accuracy even further. Recall that a voice print analyzer determines the identity of persons speaking, a feature that can be enhanced through the addition of a voice stress analyzer, measuring the level of stress in the voice through low frequency stress components (also indicative of lying). Other aspects of speech, such as heightened inflection, also indicate insincerity in the form of jocularity or facetiousness. Facial expressions further betray underlying emotions, the unconscious use of distinctive sets of facial muscles allowing for a reliable style of emotional analysis. A camera set-up for determining facial expressions can further be extended (with a telephoto lens) to monitor eye pupil size (an autonomic function that further betrays underlying emotion). Body posture, body synchrony, and associated mannerisms further prove informative in such a detection mode.

[0166] The further addition of direct contact parameters, such as those conventionally monitored in the traditional lie detector set up could also prove extremely useful in a diagnostic sense. These include a galvanic skin response detector (GSR), a blood pressure cuff, along with a chest strain gage (for measuring the rate and depth of breathing). All of these non-verbal detection parameters are lumped together in FIG. 13 as step 31, adding an additional degree of confidence to the power pyramid definition matching procedure, as coordinated through specialized subroutines within the MCU.

[0167] Summary, Ramifications, and Scope

[0168] In summary, the preferred version of the IIALA is the coordination of two basic expert system components; namely, a power pyramid definition matching procedure expert system, and the master control unit expert system. The latter coordinates the system over time, as well as overseeing the generation of computer responses (allowing for a simulation of AI). At the heart of this overall system are the listings of power pyramid definitions, as documented in the 32 pages of tables. These power pyramid definitions are indispensable to the heart of this system, the underlying knowledge base without which this process would have remained impossible.

[0169] Although the description above contains many specificities, these should not be construed as limiting the scope of the invention, but as merely providing illustrations of some of the presently preferred embodiments of this invention. For example, the passive monitoring mode exhibits applications to monitoring and surveillance functions, when a simple “sound the alarm” response is sufficient. These include the role of night watchman, security guard, or child monitor. The passive monitoring mode also exhibits the potential for applications as a smart surveillance tape, where a synopsis of the content of an ongoing verbal interaction is instantaneously available.

[0170] The optional active monitoring mode allows the use of simple yes or no questions by the computer for clarifying ambiguities, allowing for the more versatile roles of screening/interviewer, receptionist, or public-relations specialist. The preferred true AI mode makes no such restrictions on its response repertoire, excelling in applications where a convincing simulation of motivational language function is specified, roles including personal assistant or personal digital companion. The further use of non-verbal cues in the decoding and matching procedure adds an additional level of confidence in the process. The addition of various expert systems add-on programs extends these digital applications to professional fields where expert systems have been devised.

[0171] Although such applications typically refer to active verbal input, such functions also extend to written language input as well. The vast accumulated tradition of literature is similarly amenable to decoding within such a system, allowing for a master data-base of the great literature of the world: an eminently searchable format classified by detailed affective content.

[0172] Accordingly, the scope of this invention should be determined not by the embodiments illustrated, but by the appended claims and their legal equivalents.