5 Coding simulation circuits during symbolic interaction

What enables the speaker and listener in conversation to avoid misunderstanding prior to external feedback from one another? This paper, presented at a cybernetics meeting in Namur 1973, offers this reply: By virtue of internal alter-oriented simulation circuits ego is capable of alter-oriented monitoring and self-regulation of own acts through simulation of reverse alter-processes. The speaker regulates what he is about to say by simulating the reverse process in the listener; i.e. egos' encoding of what is (to be) said is regulated by predictory simulation of alter's decoding. The listener regulates his processing by simulating the reverse process in the speaker, i.e. ego's decoding of what has been heard is regulated by postdictory simulation of alter's encoding. This model is based on the treatise quoted in the previous essay (no.4, this volume), but applied here to sensory-motor operations also in a more general sense: the ego-oriented TOTE-unit specified by Miller, Galanter and Pribram (1960) is proposed extended to include alter-oriented SOS-circuits evoked in interaction situations.

Coding regulations through simulations
The human actor participating in symbolic interaction appears to be able to employ internal correction signals warning him of the danger of misunderstanding or being misunderstood by his coactor prior to overt production, and prior to reception of any correction signals from the coactor. Apparently, internal feedback circuits are operant in human language behaviour (Bråten 1973ab, Lenneberg 1967, Liberman 1957, Mead 1934, Miller & Chomsky 1963, Mysak 1971, Quillian 1968, Rommetveit 1972ab). Thus the question regarding the criteria used -- or in case of a machine, that may economically be used -- in internal encoding and decoding regulation becomes critical. Concerning regulation of decoding, Miller and Chomsky suggest for their sentence-recognition device that reconstruction of the generation of the input signal may serve as a test for understanding the input. This is analogous to the principle described in Liberman's (1957) report on motor theory of speech perception, and the speech analysis-by-synthesis theory of Halle and Stevens (1968). Quillian (1968) discards such a comparison test on understanding as unnecessary when a single message content is assumed as underlying both production and understanding; construction of that content is regarded as sufficient. This is also assumed by Rommetveit (1972b) provided that the speaker and the listener have established a shared world and a contract of complementarity, involving anticipatory decoding on the part of the producer.
The latter point is consistent with Mead's (1934) notion of anticipatory response as part of his symbolic interactional frame, which easily allows for expression and extension in a cybernetic-oriented language, as done by Bråten (1973b). According to this frame, man's symbolic activity cannot be understood or described as significant acts as long as he is conceived as a monad locked in a cell. The significant acts are made possible by ego's capacity for alter-oriented regulation of own acts by his assuming the role or attitude of the other, i.e. employing a coactor-model.
Mead's notion of anticipatory response may be adopted and extended as part of a theory for encoding and decoding regulation, containing the postulate put forward by Bråten (1973a) to the effect that the actor regulates his coding activity through simulation of reverse coactor's coding processes, i.e. through covert experimentation with his coactor-model:

P1) During the execution of a symbolic production program the successive encoding of intended content (Cp) into a selected sign form compound (Fp) is regulated on the basis of feedback from simulated predictory decoding (Csim) of the tentatively selected forms planned for production.

P2) During the execution of a symbolic reception program the successive decoding of a received, identified utterance (Fr) into an interpreted content (Cr) is regulated on the basis of feedback from simulated postdictory encoding (Fsim) of the emerging content tentatively assumed as interpretation.

According to P1 and P2 the actor regulates his encoding through covert predictory simulation of coactor's decoding, and regulates his decoding activity through covert postdictory simulation of coactor's encoding. Such simulations are dependent upon the actor's capacity for utilizing information that may be yielded by a situational definition, including a model of his coactor, allowing for covert for experimentation.
Symbolic interaction is conceived as occurring through interconnected execution of symbolic action programs, which have a produced utterance as output (production program) or take an utterance as input (reception program). Thus P1 and P2 refer to a subset of the set of processes activated in the execution of such action programs. The encoding regulation referred to by P1 utilizes as negative feedback the discrepancy between Cp and Csim. The decoding regulation referred to by P2 utilizes as negative feedback information on the discrepancy between Fr and Fsim.

Testable implications
Real world systems that show symptoms of verbal sign encoding or decoding impairments may be used to evaluate P1 and P2 through their implication about such systems. As usage of decoding mechanisms during regulated encoding is assumed by P1, it follows that

h1) whenever clear symptoms of decoding aphasia are found, then symptoms of encoding regulation impairments will also be found.

As usage of encoding mechanisms during regulated decoding is assumed by P2, it follows that

h2) whenever clear symptoms of encoding aphasia are found, then symptoms of decoding regulation impairments will also be found.

Pure cases of decoding aphasia are thereby prohibited by P1 through h1, and pure cases of encoding aphasia are prohibited by P2 through h2. Thus a single reported clinical case of a pure form would suffice to falsify either of the propositions. So far, unsystematic search for, and retrieval of, reports and descriptions of reports have not revealed to the author any clear-cut case that refutes P1. As to the tenability of P2, Fog and Herman (1967) report on a case of pure amnesic aphasia, characterized by difficulties in mobilizing the adequate world symbol. The patient may reject proposals that do not match his intended conceptual content, and show relief when recognizing the correct suggestion among a set of encoding suggestions. However, this form involves failure in creating any output of the symbol-selection process, and not in the control or regulation of such an output, and is thus not a clear-cut falsification case of P2 through h2.
Lenneberg (1967) points out that studies of language development in defective children reveal that knowledge of language may be established in the complete absence of capacities for language responses. Productive ability does not appear to be critical for language perception and development. This generalization seems to falsify P2. However, the studies he refers to concern cases of children with inborn disability to coordinate their muscles of the vocal tract sufficiently to produce intelligible speech, and need not exclude the possibility of inner speech. However, the observable fact that a child of less than 2 years old is capable of responding adequately to language sentences without being able to produce such sentences may constitute a basis for refutation of P2. This has lead the author to a less strict formulation, in which decoding regulation through encoding simulation is seen as a possible, but not necessary, means of adaptation during symbolic interaction.
Lenneberg (1967) shows that such clinical occurrences as acquisition of language-understanding in the child with organic disability to speak can be explained by "the assumption of a single machine that runs its own course like an automation", and "serves at the same as receiver, interpreter, and producer for language messages". We may add, according to the above modification, that his machine may be conceived as equipped with warning bells, allowing for activation of coactor-oriented regulation circuits and thereby interfering with the otherwise ego-oriented and automated course during symbolic interaction.

The TOT(SOS)E scheme
The scheme embedded in the sociosemantic structure outlined above may be conceived as an extension of the TOTE principle (for Test, Operation, Test,...,Exit) introduced by Miller et al. (1960) and applied in language behaviour modelling by several authors (Miller & Chomsky 1963; Pribram 1971; Rommetveit 1972b). In modified form with reference to grosser level of resolution, the principle allows for inclusion of so-called SOS-circuits (for Simulation-test, Operation,..., Simulation-test), involving parallel feed-forward processes in line with the approach of Pribram (1971) to the neurophysiological organization of the brain. In the present context we are dealing with the kind of actor's behaviour programs that require participation by a coactor in order to be completed, and such that execution of some operations (such a encoding) as part of the action program is dependent upon some reverse coactor operation (such as decoding) in order to satisfy criteria applying to the interaction program execution. The SOS-circuit allows for actor's regulation of his operation through his simulation of the reverse coactor-operation. In case of pure ego-oriented activity, these kinds of circuits need not be activated (see Fig. 5.1).

Without specifying the kind of operation involved, let X denote the input to the operation O, producing some tentative output Y' as candidate for exit E, after some initial ego-oriented automated test T has been satisfied, employing one or several negative feedback loops. Now, if the operation in question is part of a coactor-oriented activity and, thus, of a coactor-dependent interaction program, the SOS-circuit may be activated. It takes Y' as input to a simulated reverse coactor-operation with Xsim as simulation output. This simulation test involves comparison of the initial input X and Xsim. In case of unacceptable discrepancy the operation O is re-activated, producing another tentative output Y". This may in turn be subjected to simulation testing until some candidate for exit is allowed through exit.
While allowing for a larger domain, the TOT(SOS)E scheme is in the present context restricted to coding regulation. In case of encoding regulation, simulation of coactor's decoding is assumed as the SOS-part. In case of decoding regulation, activation of the SOS-circuit means carrying out regulation of reconstructional simulation of coactor's encoding.

Encoding regulation through simulation
Let us now consider the TOT(SOS)E scheme as part of the set of operations building up a production program, including activation of coactor-in-game model, content input generation, encoding operations and tests, and sign production. In this context the regulation circuit is based on predictory simulated decoding, as expressed by P1; the symbol X denotes content input and Y denotes encoded forms allowed through exit to sign production. Thus, generated content serves as input to the automatic, ego-oriented encoding operations and tests, resulting in a final (Y) or tentative selection (Y') of sign forms. The alter-oriented SOS circuit accepts the tentative forms as input to simulated coactor-decoding, and compare the simulated interpretation (Xsim) with the actual, intended content (X). (This would correspond to the posed question: "Will my coactor in our game decode (understand, interpret) the forms planned for production in congruence with my intention?"). In the case of unacceptable discrepancy (in terms of X-Xsim), encoding operations are re-activated, followed by another simulation test, etc... until some final encoding output Y is allowed through exit to sign production. The TOT(SOS)E encoding circuits are thereby completed upon satisfactory outcome of the following comparison test:

(i) Does the output of simulated coactor's decoding match the content input to own encoding? This provides an informational basis for coactor-oriented encoding regulation.

Decoding regulation through simulation
When the scheme is interpreted as illustrating the proposition P2 regarding the possibility of decoding regulation through postdictory encoding simulation as part of a reception program, X denotes identified sign forms as input to decoding operations and tests, and Y' denotes some tentative interpretation. In the run of a normal ego-oriented reception program, the output of the decoding of the identified sign input may be allowed through exit as the selected interpretation Y. In the case of signals warning about the inadequacy of the tentatively selected direction of interpretation, the SOS part may be activated, involving simulated postdictory encoding of the tentative interpretation. (This corresponds to the question: "Would my coactor have expressed my present understanding of his expressed forms by those forms which he actually used?") Thus, the comparison test involved concerns this relation:

(ii) Does the output of simulated coactor's encoding match the sign input to own decoding? This provides a basis for coactor-oriented decoding regulation.

In case of unsatisfactory discrepancy (in terms of X-Xsim) the decoding operations may be (1) re-activated or (2) the identification of sign form input may be changed. Selection of solution mode (1) or (2) is here conceived as a function of the degree of discrepancy between input and simulation output and the number of trials preceding the comparison test, so that the probability of selection of the more drastic mode (2) increases with the degree of discrepancy and the number of trials.
A similar set of solutions are open to the system in case of unsolved discrepancy during execution of a production program with respect to (i), i.e. unacceptable discrepancy between X and Xsim as encoding input and simulated coactor-decoding output may lead to (1) re-encoding of the given content, or (2) the more drastic solution of changing intended content. (The ultimate drastic mode of withdrawal from the communication game is here disregarded).

Utilization of external feedback
Simulated coactor-decoding provides not only for internal feedback in the regulation of own encoding, but also feedforward basis for evaluating actual coactor reaction, and thus for adjusted situational definition, including adaptation of the actor's coactor-model. Actual coactor feedback (as identified sigh forms) may be compared with expected coactor-reaction upon the simulated coactor decoding of own produced forms. In case of unacceptable discrepancy between expected reaction (on the basis of simulation) and actual coactor reaction, the present coactor-in-game model may be adjusted or replaced in an effort to establish a more valid situational definition, for use in the current reception program and the next production program. Thus this kind of test, utilizing actual coactor feedback and the output of simulated decoding (from the SOS circuit of the last production program execution), involves the following comparison:

(iii) Does simulated coactor reaction (upon simulated coactor decoding) match actual coactor reaction (identified sign input)? This provides the basis for adjustment or shift of coactor-model as part of situational re-redefinition.
Needless to say, none of the above tests, (i), (ii), (iii), need be activated when perfect alter-orientation on the part of the coactor, or completely shared world, code and game definition are assumed by the actor, or when he does not care about the symbolic interaction outcome.
1 Read at the 7th International Congress on Cybernetics, Namur 1973, published in The Proceedings/Actes 7e Congres International de Cybernétique, Association Internationale de Cybernétique, Namur 1974, 327-336 (introductory and concluding parts (pp. 327, 333-36) have been omitted in this reprint).