Concept of Agent

14.1 Concept of Agent

It is difficult to point to the psychological theory or theory of mind which has influ- enced the development of the concept of agent in Artificial Intelligence the most. Surely, we can see in this concept the behavioral idea of explaining the behavior of organisms in terms of stimulus-response. We can also find the influence of the

Piaget 2 constructivist theory of cognitive development, especially mechanisms of assimilation and accommodation 3 of concepts. Some authorities in AI claim that functionalism and the Dennett approach to intentionality 4 have had an impact on this model as well [256]. Neglecting some differences among notions of agent in the literature, we define it as a system which influences its environment in an autonomous way on the basis of perceived information in order to achieve required goals. The expression autonomous plays a key role here and means that an intentional influence on the environment should be based on the experience an agent gained itself. In other words, an agent should have an ability of self-learning and it should make use of this ability to cognize an unknown environment. Let us explain this concept with the help of the following example.

Let us return to our example of searching for a route to the exit in a labyrinth, which has been introduced in Chap. 4 . Let us consider a certain world of labyrinth paths , which is shown in Fig. 14.1 a. As one can see this world consists of paths which run from north (N) to south (S) and sometimes turn east (E). Then, we reach a corridor which runs to the west (W) to the exit. (On the east side of the map there are

a lot of similar paths and there is no exit on this side.) In this world, at the beginning of each path we place an agent. We expect it to find a route to the exit. Of course, since the agent is autonomous, we do not deliver any specific knowledge about this world to it. Instead, we assume it has the ability to perceive, basic meta-knowledge, and the ability to move.

Let us begin with defining the ability to perceive. Let us assume that the agent perceives the world one step ahead and it can recognize an empty space or a wall. Additionally, the agent knows the direction (N-E-S-W) it is moving (see Fig. 14.1 b), i.e., it has an internal compass. The ability to move is defined with the following operations: go one step ahead, go one step back, turn right, turn left.

The meta-knowledge of the agent is modeled with the help of rules of the form introduced in Chap. 9 . The rules define basic principles which allow the agent to exit the labyrinth. For example, we can define the following rules: IF I see an empty space THEN I go one step ahead, IF I see a wall THEN I turn right.

2 Jean Piaget—a professor of psychology at the Rousseau Institute in Geneva. He is known primarily for his theory of cognitive development and epistemological research with children.

3 Information perceived by humans is assimilated according to pre-existing cognitive schemas. However, if this is not possible, because the information perceived does not fit these schemas, they

are altered to take into account the new experience, which is called accommodation. 4 These theories are introduced in the next part of the book.

14.1 Concept of Agent 205

(a)

(b)

(c)

Fig. 14.1 a An exemplary world of labyrinths, b possible steps of an agent, c an example of a route to the exit

We have assumed above that the agent is autonomous. Therefore, we should give it cognitive abilities that can be used for constructing knowledge schemas. (However, we should not give these schemas to it.) For example, we can require that the agent should gain knowledge defined with the rules presented above on its own, i.e., via experiments made in its environment. In such a case, we might define the agent as a

child at the age of developing sensory-motor abilities. 5 Additionally, we should give

a body to the agent, so it experiences sensations. Then, we should define meta-rules, such as, for example, the following ones: IF I make a step ahead ∧ my body hurts THEN I go back, IF as a result of the activity A my body hurts THEN I do not repeat A.

We should also define mechanisms which allow the agent to generate specialized rules on the basis of meta-rules. Such mechanisms are modeled in the area of cognitive architectures. 6

If the agent generates the rules, such as the ones defined at the beginning of our considerations, then they will be too general from a practical point of view. They do not include any information that is specific to the world of the labyrinth the agent lives in. For example, the agent may walk to the north (although it should never do it), it may try to go east more than one step (it should not do it, because in paths running from north to south, after turning east it should go south), etc.

Thus, let us give new cognitive abilities to the agent which enable it to create schemas of a specific world. Firstly, we allow the agent to gather experience, i.e., to

5 We discuss kinesthetic intelligence in the next chapter. 6 For example, such mechanisms are implemented in cognitive architecture systems, such as Soar

and ACT-R mentioned in the first chapter.

206 14 Cognitive Architectures store paths while searching for the exit. Secondly, in order to remember only those

parts of paths which bring the agent closer to the solution, we add rules of the form: if I went along a path which ended at a cul-de-sac and I had to go back (e.g., the agent went north), then this part of my walk should be deleted, etc.

If we define such meta-rules in an adequate way, then the agent can memorize the shortest path to the exit, i.e., one without parts corresponding to wandering. For example, in case the agent starts at the point shown in Fig. 14.1 a, this is the path shown in Fig. 14.1 c. This path can be represented by the expression S 4 ES 5 W 2 . Similarly, the second path can be represented by S 2 ES 3 ES 4 W 5 and the third one by S E S 2 ES 2 ES 4 W 9 . These expressions can be treated as sentences of a formal language and they can be used for inferring a formal grammar for this language with the help of the grammatical induction algorithm introduced in Chap. 8 . Then, given the grammar we can construct a formal automaton 7 that can be used by the agent for navigating in the world of labyrinths. At the beginning of our considerations we have assumed that the agent should be autonomous . In our example this means that we have not given the agent a scheme representing the world with the help of the formal automaton, but we have given the agent the ability to construct such an automaton. Thanks to this approach, if the agent goes to another world of labyrinths, such as, for example, the rotated world

of labyrinths shown in Fig. 14.2 a, 8 it should be able to solve the problem as well. Having cognitive meta-rules, the agent is able to construct a corresponding formal automaton after some time. As a result, such an automaton would help the agent to

find the shortest path to the exit, like the one shown in Fig. 14.2 b. We can give cognitive abilities to the agent which are based on various methods of Artificial Intelligence. The exemplary agent 9 shown in Fig. 14.2 c is implemented with the help of three different models. After perceiving information concerning its environment (in this case it is an environment consisting of various devices and equipment), the agent updates its representation of this environment, which is defined as a semantic network. Then, it reasons about its situation in this environment with the help of a graph automaton (parser) in order to detect events which require its response. After recognizing the type of the event, the agent infers an adequate action with the help of a rule base. Proposed actions are sent to a steering command generator , which is used to influence the environment. The agent has the ability to

learn by inferring a formal grammar, 10 which is the knowledge base for a graph automaton (parser).

7 As is discussed in Chap. 8 , as well. 8 Our agent should be able to go to any world of labyrinths, assuming paths in such a world are characterized by some regularities (i.e., they can be described by some principles). 9 The example is discussed in the paper Flasi´nski M.: Automata-based multi-agent model as a tool for constructing real-time intelligent control systems. Lecture Notes in Artificial Intelligence 2296 (2002), 103–110.

10 This is an ETPL(k) graph grammar introduced in Chap. 8 and formally defined in Appendix E.

14.2 Multi-agent Systems 207

(a)

(b)

E 4 NE 5 S 2

(c)

Input

information (grammar

Steering command,

message AGENT

Fig. 14.2 a A rotated world of labyrinths, b the route to the exit in the rotated world, c an example of an agent structure

Dokumen yang terkait

Hubungan pH dan Viskositas Saliva terhadap Indeks DMF-T pada Siswa-siswi Sekolah Dasar Baletbaru I dan Baletbaru II Sukowono Jember (Relationship between Salivary pH and Viscosity to DMF-T Index of Pupils in Baletbaru I and Baletbaru II Elementary School)

0 46 5

Institutional Change and its Effect to Performance of Water Usage Assocition in Irrigation Water Managements

0 21 7

The Effectiveness of Computer-Assisted Language Learning in Teaching Past Tense to the Tenth Grade Students of SMAN 5 Tangerang Selatan

4 116 138

the Effectiveness of songs to increase students' vocabuloary at second grade students' of SMP Al Huda JAkarta

3 29 100

The effectiveness of classroom debate to improve students' speaking skilll (a quasi-experimental study at the elevent year student of SMAN 3 south Tangerang)

1 33 122

Kerjasama ASEAN-China melalui ASEAN-China cooperative response to dangerous drugs (ACCORD) dalam menanggulangi perdagangan di Segitiga Emas

2 36 164

The Effect of 95% Ethanol Extract of Javanese Long Pepper (Piper retrofractum Vahl.) to Total Cholesterol and Triglyceride Levels in Male Sprague Dawley Rats (Rattus novergicus) Administrated by High Fat Diet

2 21 50

Factors Related to Somatosensory Amplification of Patients with Epigas- tric Pain

0 0 15

The Concept and Value of the Teaching of Karma Yoga According to the Bhagavadgita Book

0 0 9

Pemanfaatan Permainan Tradisional sebagai Media Pembelajaran Anak Usia Dini untuk Mengembangkan Aspek Moral dan Bahasa Anak Utilization of Traditional Games as Media Learning Early Childhood to Develop Aspects of Moral and Language Children Irfan Haris

0 0 11