Intentional Cognitive Architectures

Intentionality and Intentional Primitives

In Grice's definition, an interaction between two agents is an instance of communication when one of them entertains a certain, precisely definable set of mental states [namely, intentions (1), (2) and (3) above] and the other entertains, as a consequence of the first agent's relevant behav­ior, a certain other [namely, the belief that she entertains intentions (1), (2), and (3)].

A framework of this kind, called an Intentional one and founded on the concept of mental states, is unusual in cognitive psychology and neuropsychology; nonetheless, it has become the common choice of most current theories of pragmatics, whether they take a philosophical, logi­cal, computational, or psychological perspective.2 Remarkably, it has also been widely adopted in recent years for research on high-level agency in cognitive science, artificial intelligence, and distributed artificial intelligence, where talk of BDI (belief-desire-intention) architectures is cur­rently standard (e.g., Rao & Georgeff 1992). This is particularly telling, since studies in this

area used to adopt instead a more classic information-processing stance, until it became clear that planning and deliberation are better treated in a mental states framework (Bratman 1987; Cohen & Levesque 1990a; Pollack 1990).

2 Let me remind that the word intentional and its derivatives have two meanings. One corresponds, roughly, to semantic, or representational, and applies therefore to any mental state endowed with content, like a percep­tion, a belief, a fear, or a desire. The other refers instead to a specific class of such states, namely, intentions, that are relative to action planning and execution. For the sake of clarity, I capitalize the initial "I" when using the former acceptance.

In general, Intentional approaches prove the most fitting for the development of theories aimed at capturing the complexity of an agent's interactions with the real world. Concepts like plan, expectation, or preference or whole stories like Ann went to the movies because she was bored with staying at home are hardly reducible to the coding, the processing, and the storing of information. They build instead on a conception of the mind as representing a whole subjective situation, whose every detail only makes sense within the broader picture, and on a view of the agent's decisions and actions as conscious, purposeful movements within this complex land­scape in search of an improvement to the current situation.

The same holds for social interactions: concepts like assert/reply and promise/accept or coop­eration (or lack thereof) and teamwork as well as the umbrella episode we have seen above only make sense if the mental states of the agents involved are taken into account — witness also the difficulty to define the difference between, say, beg, borrow, and steal in a behavioral or an in­formation-processing framework.

Concepts and stories like these are instead more naturally decomposed into simpler situa­tions, decisions and actions of the same general type, like Ann picked up the newspaper to see what movies were on that day. Thus, Intentional theories build ultimately on what may be called Intentional primitives, that is, the types of mental states, of contents for them, and of transitions from one state to another that are characteristic of a certain agent, or species of agents, under certain circumstances.

It has been argued against Intentional terminology that it belongs to "folk psychology" and should therefore be banned from the scientific study of cognition (Churchland 1984, 1988; Churchland 1986; Stich 1983). Concepts like belief or intention, though, cannot render a theory of psychology more "folk" than the concepts of force, field, or reaction can with theories of physics or chemistry. While the everyday acceptation of all these terms may often be vague, each also has a technical meaning (possibly rooted somehow in the "folk" one) that can be de­fined with better

precision and adopted in scientific theories in psychology as in physics or chemistry.

However, there is little consensus in cognitive science as to the precise meaning of Intentional terms. My acceptation is that mental states are conscious representations of the world as it, from the agent's subjective point of view, is, was, or could be. For an agent to entertain a certain mental state is for her to be in a certain semantic relation to the world; the various types of mental states that she may entertain are therefore defined according to the different types of semantic relations that she may have with the world (Searle 1983). Theories may thus be formulated where, for example, beliefs encode the agent's knowledge of the situation at a given time, desires encode her potential goals, intentions encode her actual goals, and so on.

The entities that an agent's mind is sensitive to and able to represent — in Intentional terms, the possible contents of her mental states — make up her subjective ontology (Tirassa, Carassa & Geminiani in press). This may comprise different types of objects, relations, events, and ac­tions. The subjective ontology of a highly sophisticated social species like ours will also com­prise social objects (individual agents, groups, etc.) along with their features (beliefs, inten­tions, etc.) and the relations, events, and actions that they may participate in (staying at home, going to the movies, etc.).

Методика «Время года»

Цель: изучение наглядно-образного мышления детей.

Проведение исследования: Ребенку показывают рисунок и просят, внимательно посмотрев на этот рисунок, сказать, какое время года изображено на каждой части данного рисунка. За отведенное на выполнение этого задания временя – 2 мин. – ребенок должен не только назвать соответствующее время года, но и обосновать свое мнение о нем, т.е. объяснить, почему он так думает, указать те признаки, которые по его мнению, свидетельствуют о том, что на данной части рисунка показано именно это, а не какое-либо иное время года.

Оценка результатов:

10 баллов – за отведенное время ребенок правильно назвал и связал все картинки с временами года, указав на каждой из них не менее двух признаков, свидетельствующих о том, что на картинке изображено именно данное время года (всего не менее 8 признаков по всем картинкам).

8-9 баллов – ребенок правильно назвал и связал с нужными временами года все картинки, указав при этом 5-7 признаков, подтверждающих его мнение, на всех картинках, вместе взятых.

6-7 баллов – ребенок правильно определил на всех картинках времена года, но указал только 3-4 признака, подтверждающих его мнение.

4-5 баллов – ребенок правильно определил время года только на одной - двух картинках из 4-х и указал только 1-2 признака в подтверждение своего мнения.

0-3 балла – ребенок не смог правильно определить ни одного времени года и не назвал точно ни одного признака.

Выводы об уровне развития.

10 баллов - очень высокий; 8-9 баллов – высокий; 6-7 баллов - средний

4 - 5 баллов – низкий 0-3 балла - очень низкий

 

 

Communicative Competence

Domain-General vs Domain-Specific Views of Cognition

A mental phenomenon is called a competence (e.g., Chomsky 1980) when it is conceived of as yielded by the functioning of a mental organ, that is, of an innately dedicated part of the mind/brain. Human language has been considered a prototypical example for 4 decades now, at least as far as grammar is concerned. In recent years, several theories have been proposed that posit the existence of other faculties, from face recognition to deontic reasoning and social exchange, thus giving rise to a lively debate between domain-general and domain-specific views of cognition.

The domain-general stance postulates the existence of a core, or central, cognitive system, whose functioning is characterized by general principles of cognition that are independent of whether their specific content regards communication or, say, physical causality. These princi­ples hold across the different domains where the mind operates and may be formulated, e.g., as general laws of cognition (in the fashion of Sperber & Wilson's, 1986, principle of relevance, or of Newell's, 1990, proposal of SOAR) or abstract inferential schemes (like those found in Newell & Simon's, 1972, or Johnson-Laird's, 1983, theories of human reasoning). Fodor's thesis of the language of thought (1975, 1983; Fodor & Pylyshyn 1989) is paradigmatic of this perspective.

The other side of the controversy holds instead that the mind operates with different princi­ples in each domain because the ways of interpreting the world and acting in it that constitute adaptive cognition in one domain are unlikely to keep their utility in another so that no "adaptiv-ity-preserving" generalization is guaranteed to hold across domains. The search for general laws of cognition is therefore doomed to fail unless such laws are formulated so vaguely as to lose much of their scientific interest. The study of the architecture of cognition should consequently follow the natural decomposition of the mind/brain into specialized organs rather than an ab­stract, unwarranted decomposition into allegedly general-purpose perceptual, storing, reason­ing, and acting subsystems.

There is no room here to provide a review of the debate; suffice it to say that, in my view, theoretical considerations (Cosmides& Tooby 1994a; James 1890; Minsky 1985; Pinker 1997) as well as empirical findings in evolutionary psychology (Barkow, Cosmides & Tooby eds. 1992; Cosmides & Tooby 1994b) and in developmental psychology (Hirschfeld& Gelman eds. 1994; Karmiloff-Smith 1992; Spelke 1994) strongly suggest a decentralized, faculty-based view of cognition.

A domain-specific perspective is also in better agreement with current advancements in neu-rosciences, providing increasing evidence against the idea that vast areas of the brain may be dedicated to the sort of unspecific, equipotential processing that is definitional of a centralized architecture.

Let us define a domain as a specific type of interaction between an agent and its subjective environment, or, circularly, as the class of problems that admit of a certain class of solutions. Communication seems a natural candidate for this definition. The idea of a distinct communica­tive competence, however, has been taken into little account: implicitly or (more seldom) ex­plicitly, and with very few exceptions, most researchers view communication as just an aspect of the functioning of a general-purpose cognitive system (e.g., Cohen & Levesque 1990b; Kasher 1991a, 1991b; Sperber & Wilson 1986; Wilson & Sperber 1991).

It is the aim of the rest of this section to argue in favor of a competence view of communica­tion. Communicative competence does not have to be similar in nature to the Universal Grammar that has been hypothesized for language. Rather, the problem of whether communi­cation should be viewed as a faculty may be reframed within the metatheoretical perspective outlined in the previous sections as the problem of whether it requires a dedicated cognitive dy­namic. The issue requires a little bit of technical discussion.

Задание на составление сказки

Тема: Составление сказки с использованием серии сюжетных картинок по сказке «У страха глаза велики».

Цель: изучение уровня развития речи детей.

Оборудование: серия сюжетных картинок по сказке «У страха глаза велики».

Ход занятия:

Воспитатель сообщает детям, что сегодня дети будут составлять сказку по картинке. Просит вспомнить, как обычно начинаются сказки и как заканчиваются. Предлагает назвать известные им сказки.

Затем дети называют свои любимые сказки и воспитатель предлагает детям вспомнить известную им сказку «У страха глаза велики».

Воспитатель начинает рассказывать:

По синю морю, корабль бежит,

Серый волк на носу стоит,

А медведь паруса крепит.

Заюшка кораблик за веревку ведет,

Лисичка из-за кустика хитро глядит:

Как бы зайку украсть.

Как бы веревку сорвать.

Это еще не сказка, а присказка,

а сказка вся впереди.

Воспитатель прерывается на этом и говорит:

«Вы узнали сказку? Давайте сегодня попробуем рассказать ее сами, но изменим некоторые моменты, т.е. составим новую сказку. Картинки помогут вам в этом».