Computational Explanation in BIC(A/S)

Antonio Lieto

University of Turin, Dipartimento di Informatica, ItalyICAR - CNR, Palermo, Italy

http://www.di.unito.it/~lieto/

Fierces on BICA, International School on Biologically Inspired Cognitive Architectures

Moscow, Russia, 21-24 April 2016

From Human to Artificial Cognition (and back)

2

Inspiration

Explanation

Lieto and Radicioni, Cognitive Systems Research, 2016

Research Questions

When a biologically inspired computational system/architecture has an explanatory power w.r.t. the natural system taken as source of inspiration ?

Which are the requirements to consider in order to design a computational model of cognition with an explanatory power ?

3

Outline

- Cognitive AI Paradigms: some methodological and technical considerations.

- Functionalist vs Structuralist Approach.

- Case study on Knowledge Level in Cognitive Architectures.

4

Cognitive AI

Attention to the heuristics-based solutions adopted by humans (e.g. Gigerenzer & Todd, 1999) for combinatorial problems (“bounded rationality heuristics”).

Heuristics realize/implement some cognitive functions and are responsible of the macroscopic external behaviour of an agent.

5

“Natural/Cognitive” Inspiration and AI

Early AI

Cognitive Inspiration for the Design of “Intelligent Systems”

M. Minsky

R. Shank

Modern AI

“Intelligence” in terms of optimality of a performance

(narrow tasks)

mid‘80s

A. Newell

H. Simon

e.g. IBM Watson…

N. Wiener

A focus shift in AI

Vision the early days of AI: “Understanding and reproducing, in computational systems, the full range of intelligent behavior observed in humans” (P. Langley, 2012).

This view was abandoned. Why?

- Emphasis on quantitative results and metrics of performance: (“machine intelligence”: achieving results and optimize them !)

- Renewed attention since “The gap between natural and artificial systems is still enormous” (A. Sloman, AIC 2014).

7

2 Main Perspectives

“Cognitive Systems” (Brachman and Lemnios, 2002): “designs, constructs, and studies computational artifacts that exhibit the full range of human intelligence”. [Cognitivist approach, Vernon 2014].

“Nouvelle AI” (e.g. Parallel Distributed Processing (Rumhelarth and McLelland, 1986) based on bio-plausibility modelling techniques allowing the functional reproduction of heuristics in artificial systems (neglecting the physical and chemical details). [Emergent approach, Vernon 2014].

8

Type 1/Type 2 features

9

Cognitivism Nouvelle AI

Focus on high level cognitive functions Main focus only on perception

Assuming structured representations (physical symbol system, Simon and Newell, 1976)

Assuming unstructured representation (e.g. such as neural networks etc.) and also integration with symbolic approaches.

Architectural Perspective (integration and interaction of all cognitive functions

System perspective (not necessary to consider a whole architectural perspective).

Inspiration from human cognition (heuristic-driven approach)

Bio-inspired computing, bottom-up approach (for learning etc.).

A Matter of Levels

• Both the “classical” and “novuelle” approach can realize, in principio, “cognitive artificial systems” or “artificial models of cognition” provided that their models operate at the “right” level of description.

• A debated problem in AI and Cognitive Science regards the legitimate level of descriptions of such models (and therefore their explanatory power).

Functionalist vs Structuralist Models 10

Functionalism • Functionalism (introduced by H. Putnam) postulates a weak

equivalence between cognitive processes and AI procedures.

• AI procedures have the functional role (“function as”) human cognitive procedures.

• Multiple realizability (cognitive functions can be implemented in different ways).

• Equivalence on the functional macroscopic properties of a given intelligent behaviour (based on the same input-output specification).

• This should produce predictive models (given an input and a set of procedures functionally equivalent to what is performed by cognitive processes then one can predict a given output). 11

Problems with Functionalism

• If the equivalence is so weak it is not possible to interpret the results of a system (e.g. interpretation of the system failure…).

• A pure functionalist model (posed without structural constraints) is a black box where a predictive model with the same output of a cognitive process can be obtained with no explanatory power.

12

Birds and Jets

- Both a Bird and a Jet can fly but a jet is not a good explanatory model of a bird since its flights mechanisms are different from the mechanism of bird.

- Purely functional models/systems are not “computational models of cognition” (they have no explanatory power w.r.t. the natural system taken as source of inspiration). 13

Structuralism

• Strong equivalence between cognitive processes and AI procedures (Milkowski, 2013).

• Focus not only on the functional organization of the processes but also on the human-likeliness of a model (bio-psychologically plausibility).

14

Wiener’s “Paradox”

“The best material model of a cat is another or possibly the same cat”

- Difficulty of realizing models of a given natural system.

- Need of proxy-models (i.e. good approximations) 15

A Design ProblemZ.Pylyshyn (’79): “if we do not formulate any restriction about a model we obtain the functionalism of a Turing machine. If we apply all the possible restrictions we reproduce a whole human being”

• Need for looking at a descriptive level on which to enforce the constraints in order to carry out a human-like computation.

• A design perspective: between the explanatory level of functionalism (based on the macroscopic stimulus-response relationship) and the mycroscopic one of fully structured models (reductionist materialism) we have, in the middle, a lot of possible structural models. 16

Many Structural Models Both the presented AI approaches may build structural models of cognition at different levels of details (having an empirical adequacy => Paul Verschure’s yesterday talk).

17

Cognitive Function (NL Understanding)

Cognitive Processes Neural Structures

Sintax MorphologyLexical Processing…

Biological Plausibility of Processes

Cognitive Plausibilityof the Processes

1:N 1:N

Many Structural Models Both the presented AI approaches may build structural models of cognition at different levels of details (having an empirical adequacy => Paul Verschure’s yesterday talk).

18

Cognitive Function (NL Understanding)

Cognitive Processes Neural Structures

Sintax MorphologyLexical Processing…

Bio-Physical Plausibilityof the Processes

Cognitive Plausibilityof the Processes

Cognitivism Emergent AI

Take home message (part 1)

• Cognitive Artificial Models (BICA) have an explanatory power only if they are structurally valid models (realizable in different ways and empirically adequate).

• Cognitive Artificial Systems built with this design perspective have an explanatory role for the theory they implement and the “computational experiment” can provide results useful for refining of rethinking theoretical aspects of the natural inspiring system.

Lieto, under review

Case Study: Knowledge in Humans and CAs

• Knowledge in Humans

• Knowledge Representations in some current Cognitive Architectures (CLARION, LIDA, ACT-R)

20

In Cognitive Science there were/are different contrasting theories about “how humans represent and organize the information in their mind”…This research also influenced Artificial Intelligence

21

Classical Theory – Ex.

22

TRIANGLE = Polygon with 3 corners and sides

Prototype Theory

23

Prototype Theory (Rosh E., 1975)

Category membership is not based on necessary and sufficient conditions but on typicality traits.

There are members of a category that are more typical and cognitively relevant w.r.t. others.

Ex: BIRD, {Robin, Toucan, Penguin…}

Prototypes and Prototypical Reasoning• Categories based on prototypes (Rosh,1975)• New items are compared to the prototype

atypical

typical

P

Exemplars and Exemplar-based Reasoning• Categories as composed by a list of exemplars. New

percepts are compared to known exemplars (not to Prototypes).

Heterogeneous Hypothesis

The different proposals that have been advanced can be grouped in three main classes: a) fuzzy approaches, b) probabilistic and Bayesan approaches, c) approaches based on

non-monotonic formalisms.

Different representational structures have different accessing procedures (reasoning) to their content.

Prototypes, Exemplars and other conceptual representations can co-exists and be activated in different contexts (Malt 1989).

Representations and Cognitive Mechanisms

– Conceptual structures as heterogeneous proxytypes (Lieto 2014).

A proxytype is any element of a complex representational network stored in long-term memory corresponding to a particular category that could be tokened in working memory to “go proxy” for that category (Prinz, 2002) => inspired by Barsalou (1999)

Ex. Heterogeneous Proxytypes at work

29

The different proposals that have been advanced can be grouped in three main classes: a) fuzzy approaches, b) probabilistic and Bayesan approaches, c) approaches based on non-monotonic formalisms.

from Lieto 2014, BICA

Conceptual Heterogeneity and CA

– How current CAs deal with Conceptual Heterogeneity ?

– Analysis of ACT-R, CLARION and LIDA Knowledge Level

– Some insights and suggestions

Type 1/Type 2 features

30

ACT-R (Anderson et al. 2004)

CLARION (Sun, 2006) Vector-LIDA (Franklin et al. 2014)

Concepts as chunks (symbolic structures)

Neural networks + Symbol Like representations

High dimensional vector spaces

Assuming structured representations (physical symbol system, Simon and Newell, 1976)

Assuming a dual representations

Vectors treated as symbol-like representations (e.g. compositionally via vector blending)

Sub-symbolic and Bayesian activation of chunks

Subsymbolic activation of conceptual chunks

Similarity based vectorial activation

Prototypes and Exemplars models of categorisation available in separation (extended in Lieto et al. IJCAI 2015)

Prototypes and Exemplars models of categorisation NOT available (proposed in Lieto et al. submitted to JETAI)

Prototypes and Exemplars models of categorisation NOT available (current work)

Upshots– There are structural differences (at the

process level) between the analysed architectures in dealing with a plausible model of human conceptual representation and reasoning.

–All of these architectures can in principle account with these constraints but ACT-R has currently a better explanatory model of the human representational and reasoning conceptual structures.

Computational Explanation in BICA

Antonio Lieto

University of Turin, Dipartimento di Informatica, ItalyICAR - CNR, Palermo, Italy

http://www.di.unito.it/~lieto/

Fierces on BICA, International School on Biologically Inspired Cognitive Architectures

Moscow, Russia, 21-24 April 2016

ReferencesGigerenzer, G., & Todd, P. M. (1999). Simple heuristics that make us smart. Oxford University Press, USA.

Langley, P. (2012). The cognitive systems paradigm. Advances in Cognitive Systems, 1, 3–13.

Lieto, A. "A Computational Framework for Concept Representation in Cognitive Systems and Architectures: Concepts as Heterogeneous Proxytypes" in Proceedings of BICA 2014, 5th Int. Conference of Biologically Inspired Cognitive Architectures, Boston, Massachusetts Institute of Technology (MIT), USA, 7-9 November 2014. Procedia Computer Science, Vol. 41 (2014), pp. 6-14

Lieto, A, Radicioni D.P. "From Human to Artificial Cognition and Back: New Perspectives on Cognitively Inspired AI Systems", in Cognitive Systems Research, 39, 1-3 (2016), Elsevier

Lieto, A., Daniele P. Radicioni D.P. and Rho, V. A Common-Sense Conceptual Categorization System Integrating Heterogeneous Proxytypes and the Dual Process of Reasoning". In Proceedings of the International Joint Conference on Artificial Intelligence (IJCAI), Buenos Aires, July 2015, pp. 875-881. AAAI press.

Milkowski, M. (2013). Explaining the computational mind. Mit Press.Newell, A., & Simon, H. A. (1972). Human problem solving volume 104. Prentice-Hall Englewood.

Putnam, H.: Minds and machines. In: Hook, S. (ed.) Dimensions of Mind, pp. 138–164. Macmillan Publishers, London (1960) Pylyshyn, Z.W.: Complexity and the study of artificial and human intelligence. In: Ringle, M. (ed.) Philosophical Perspectives in Artificial Intelligence, Harvester, Brighton (1979) Rosenblueth, A., Wiener, N.: The role of Models in Sciences. Phil. Sci. 12, 316–321 (1945). Vernon, D. (2014). Artificial cognitive systems: A primer. MIT Press.