Antonio Lieto

The paper presents the heterogeneous proxytypes hypothesis as a cognitively-inspired computational framework able to reconcile, in both natural and artificial systems, different theories of typicality about conceptual representation and reasoning that have been traditionally seen as incompatible. In particular, through the Dual PECCS system and its evolution, it shows how prototypes, exemplars and theory-theory like conceptual representations can be integrated in a cognitive artificial agent (thus extending its categorization capabilities) and, in addition, can provide useful insights in the context of a computationally grounded science of the mind.

Commonsense reasoning is one of the main open problems in the field of Artificial Intelligence (AI) while, on the other hand, seems to be a very intuitive and default reasoning mode in humans and other animals. In this talk, we discuss the different paradigms that have been developed in AI and Computational Cognitive Science to deal with this problem (ranging from logic-based methods, to diagrammatic-based ones). In particular, we discuss - via two different case studies concerning commonsense categorization and knowledge invention tasks - how cognitively inspired heuristics can help (both in terms of efficiency and efficacy) in the realization of intelligent artificial systems able to reason in a human-like fashion, with results comparable to human-level performances.

Commonsense reasoning is a crucial human ability employed in everyday tasks. In this talk I provide a knowledge level analysis of the main representational and reasoning problems affecting the cognitive architectures for what concerns this issue. In providing this analysis I will show, by considering some of the main cognitive architectures currently available (e.g. SOAR, ACT-R, CLARION), how one of the main problems of such architectures is represented by the fact that their knowledge representation and processing mechanisms are not sufficiently constrained with insights coming from cognitive science (Lieto 2021; Lieto, Lebiere, Oltramari, 2018). As a possible way out to such knowledge processing issues, I present the main assumptions that have led to the development of the Dual PECCS categorization system (Lieto, Radicioni, Rho 2017) and discuss some of the lessons learned and their possible implications in the design of the knowledge modules and knowledge-processing mechanisms of integrated cognitive architectures.

Book Description (Blurb): Cognitive Design for Artificial Minds explains the crucial role that human cognition research plays in the design and realization of artificial intelligence systems, illustrating the steps necessary for the design of artificial models of cognition. It bridges the gap between the theoretical, experimental and technological issues addressed in the context of AI of cognitive inspiration and computational cognitive science. Beginning with an overview of the historical, methodological and technical issues in the field of Cognitively-Inspired Artificial Intelligence, Lieto illustrates how the cognitive design approach has an important role to play in the development of intelligent AI technologies and plausible computational models of cognition. Introducing a unique perspective that draws upon Cybernetics and early AI principles, Lieto emphasizes the need for an equivalence between cognitive processes and implemented AI procedures, in order to realise biologically and cognitively inspired artificial minds. He also introduces the Minimal Cognitive Grid, a pragmatic method to rank the different degrees of biologically and cognitive accuracy of artificial systems in order project and predict their explanatory power with respect to the natural systems taken as source of inspiration. Providing a comprehensive overview of cognitive design principles in constructing artificial minds, this text will be essential reading for students and researchers of artificial intelligence and cognitive science.

Inventing novel knowledge to solve problems is a crucial, creative, mechanism employed by humans, to extend their range of action. In this talk, I will show how commonsense reasoning plays a crucial role in this respect. In particular, I will present a cognitively inspired reasoning framework for knowledge invention and creative problem solving exploiting TCL: a non-monotonic extension of a Description Logic (DL) of typicality able to combine prototypical (commonsense) descriptions of concepts in a human-like fashion. The proposed approach has been tested both in the task of goal-driven concept invention and has additionally applied within the context of serendipity-based recommendation systems. I will present the obtained results, the lessons learned, and the road ahead of this research path.

This paper presents a practical case study showing how, despite the nowadays limited collaboration between AI and Cognitive Science (CogSci), cognitive research can still have an important role in the development of novel AI technologies. After a brief historical introduction about the reasons of the divorce between AI and CogSci research agendas (happened in the mid’80s of the last century), we try to provide evidence of a renewed collaboration by showing a recent case study on a commonsense reasoning system, built by using insights from cognitive semantics.

Dynamic conceptual reframing represents a crucial mechanism employed by humans, and partially by other animal species, to generate novel knowledge used to solve complex goals. In this talk, I will present a reasoning framework for knowledge invention and creative problem solving exploiting TCL: a non-monotonic extension of a Description Logic (DL) of typicality able to combine prototypical (commonsense) descriptions of concepts in a human-like fashion [1]. The proposed approach has been tested both in the task of goal-driven concept invention [2,3] and has additionally applied within the context of serendipity-based recommendation systems [4]. I will present the obtained results, the lessons learned, and the road ahead of this research path.

In this paper we propose a computational framework aimed at extending the problem solving capabilities of cognitive artificial agents through the introduction of a novel, goal-directed, dynamic knowledge generation mechanism obtained via a non monotonic reasoning procedure. In particular, the proposed framework relies on the assumption that certain classes of problems cannot be solved by simply learning or injecting new external knowledge in the declarative memory of a cognitive artificial agent but, on the other hand, require a mechanism for the automatic and creative re-framing, or re-formulation, of the available knowledge. We show how such mechanism can be obtained trough a framework of dynamic knowledge generation that is able to tackle the problem of commonsense concept combination. In addition, we show how such a framework can be employed in the field of cognitive architectures in order to overcome situations like the impasse in SOAR by extending the possible options of its subgoaling procedures.

Persuasive technologies can adopt several strategies to change the attitudes and behaviors of their users. In this work we present some empirical results stemming from the hypothesis - firstly formulated in [3] - that there is a strong connection between some well known cognitive biases reducible to fallacious argumentative schemata and some of the most common persuasion strategies adopted within digital technologies. In particular, we will report how both framing and fallacious-reducible mechanisms are nowadays used to design web and mobile technologies in domains ranging from e-commerce [4] and news recommendations [1] to the jihadist propaganda. We will also show how and to what extent such persuasive strategies have an impact on nudging the choices of the users in digital environments.

In philosophy of language, a distinction has been proposed by Diego Marconi between two aspects of lexical competence, i.e. referential and inferential competence. The former accounts for the relation-ship of words to the world, the latter for the relationship of words among themselves. The aim of the pa-per is to offer a critical discussion of the kind of formalisms and computational techniques that can be used in Artificial Intelligence to model the two aspects of lexical competence, and of the main difficulties related to the use of these computational techniques. The first conclusion of our discussion is that the dis-tinction between inferential and referential semantics is instantiated in the literature of Artificial Intelli-gence by the distinction between symbolic and connectionist approaches. The second conclusion of our discussion is that the modelling of lexical competence needs the advent of hybrid models integrating symbolic and connectionist frameworks. Our hypothesis is that Conceptual Spaces, a framework devel-oped by Gärdenfors more than fifteen years ago, can offer a lingua franca that allows to unify and gener-alize many aspects of the representational approaches mentioned above and to integrate “inferential” (=symbolic) and “referential” (=connectionist) computational approaches on common ground.

In this paper we present a framework for the dynamic and automatic generation of novel knowledge obtained through a process of commonsense reasoning based on typicality-based concept combination. We exploit a recently introduced extension of a Description Logic of typicality able to combine prototypical descriptions of concepts in order to generate new prototypical concepts and deal with problem like the PET FISH (Osherson and Smith, 1981; Lieto & Pozzato, 2019). Intuitively, in the context of our application of this logic, the overall pipeline of our system works as follows: given a goal expressed as a set of properties, if the knowledge base does not contain a concept able to fulfill all these properties, then our system looks for two concepts to recombine in order to extend the original knowledge based satisfy the goal.

Nel campo delle scienze cognitive molti oggi condividono l’ipotesi che siano necessari differenti tipi di rappresentazioni per modellare i sistemi cognitivi sia naturali, sia artificiali. Si considerino le rappresentazioni basate su reti neurali, i formalismi simbolici e rappresentazioni analogiche quali rappresentazioni diagrammatiche o modelli mentali. Tutti questi metodi hanno successo nello spiegare e modellare alcune classi di fenomeni cognitivi, ma nessuno è in grado di rendere conto di tutti gli aspetti della cognizione. A partire da queste considerazioni, riteniamo che sistemi intelligenti e architetture cognitive possano trarre vantaggio dalla combinazione di sistemi di rappresentazione diversi. Si pone allora il problema di fare interagire rappresentazioni di natura differente in maniera cognitivamente e teoricamente fondata. La nostra ipotesi è che gli spazi concettuali possano offrire una sorta di lingua comune, che consentirebbe di integrare e generalizzare molti aspetti delle impostazioni sopra menzionate, superando i limiti delle varie proposte intese singolarmente.

We propose a nonmonotonic Description Logic of typicality able to account for the phenomenon of the combination of prototypical concepts. The proposed logic relies on the logic of typicality ALC + TR, whose semantics is based on the notion of rational closure, as well as on the distributed semantics of probabilistic Description Logics, and is equipped with a cognitive heuristic used by humans for concept composition. We first extend the logic of typicality ALC + TR by typicality inclusions of the form p :: T(C) v D, whose intuitive meaning is that “we believe with degree p about the fact that typical Cs are Ds”. As in the distributed semantics, we define different scenarios containing only some typicality inclusions, each one having a suitable probability. We then exploit such scenarios in order to ascribe typical properties to a concept C obtained as the combination of two prototypical concepts. We also show that reasoning in the proposed Description Logic is EXPTIME-complete as for the underlying standard Description Logic ALC.

In this paper I will present an analysis of the impact that the notion of “bounded rationality”, introduced by Herbert Simon in his book “Administrative Behavior”, produced in the field of Artificial Intelligence (AI). In particular, by focusing on the field of Automated Decision Making (ADM), I will show how the introduction of the cognitive dimension into the study of choice of a rational (natural) agent, indirectly determined - in the AI field - the development of a line of research aiming at the realisation of artificial systems whose decisions are based on the adoption of powerful shortcut strategies (known as heuristics) based on “satisficing” - i.e. non optimal - solutions to problem solving. I will show how the “heuristic approach” to problem solving allowed, in AI, to face problems of combinatorial complexity in real-life situations and still represents an important strategy for the design and implementation of intelligent systems.

Questo contributo si propone di fornire uno spunto di riflessione, e una breve panoramica storica, sul ruolo che le scienze cognitive hanno giocato, e possono ancora giocare, nello sviluppo dei sistemi intelligenti di nuova generazione. Illustra, inoltre, le attività recenti che l’AISC (Associazione Italiana di Scienze Cognitive, di cui gli autori sono attualmente Vice-Presidente e Presidente) sta portando avanti per lo sviluppo di linee di ricerca nell’ambito dei sistemi artificiali di inspirazione cognitiva.

In questo contributo descriviamo un sistema di creatività computazionale in grado di generare automaticamente nuovi concetti utilizzando una logica descrittiva non monotòna che integra tre ingredienti principali: una logica descrittiva della tipicalità, una estensione probabilistica basata sulla semantica distribuita nota come DISPONTE, e una euristica di ispirazione cognitiva per la combinazione di più concetti. Una delle applicazioni principali del sistema riguarda il campo della creatività computazionale e, più specificatamente, il suo utilizzo come sistema di supporto alla creatività in ambito mediale. In particolare, tale sistema è in grado di: generare nuove storie (a partire da una rappresentazione narrativa di storie pre-esistenti), generare nuovi personaggi (ad es. il nuovo “cattivo” di una serie tv o di un cartone animato) e, in generale, può essere utilizzato per proporre nuove soluzioni e format narrativi da esplorare nell’ambito dell’industria creativa.

Questo lavoro propone un confronto tra diversi strumenti utilizzabili per modellare la conoscenza di dominio in ambito didattico: le mappa concettuali, Novak e Cañas (2006), (uno strumento tradizionalmente utilizzato nelle scuole) e le ontologie computazionali (dei sistemi formali di modellazione concettuale, attualmente molto usati nei sistemi di intelligenza artificiale per le loro capacità di “ragionamento automatico”, si veda Guarino, (1995)). Nello specifico, questo articolo presenta il risultato di un un doppio esperimento sul campo condotto presso il Liceo Scientifico “Guido Parodi” di Acqui Terme in cui gruppi di studenti paragonano lo strumento della mappa concettuale e quello dell’ontologia nella risoluzione di due problemi di “misconcezione” (o errata concettualizzazione): uno indotto attraverso la consegna di appunti e materiali didattici contenenti informazioni volontariamente contraddittorie tra loro (caso che potrebbe corrispondere alla situazione in cui uno studente prende - per qualche motivo - degli appunti in modo scorretto) e l’altro legato ad una complessità concettuale intrinseca all’argomento

In the last decades a growing body of literature in Artificial Intelligence (AI) and Cognitive Science (CS) has approached the problem of narrative understanding by means of computational systems. Narrative, in fact, is an ubiquitous element in our everyday activity and the ability to generate and understand stories, and their structures, is a crucial cue of our intelligence. However, despite the fact that - from an historical standpoint - narrative (and narrative structures) have been an important topic of investigation in both these areas, a more comprehensive approach coupling them with narratology, digital humanities and literary studies was still lacking. With the aim of covering this empty space, in the last years, a multidisciplinary effort has been made in order to create an international meeting open to computer scientist, psychologists, digital humanists, linguists, narratologists etc.. This event has been named CMN (for Computational Models of Narrative) and was launched in the 2009 by the MIT scholars Mark A. Finlayson and Patrick H. Winston1.

Formal ontologies are nowadays widely considered a standard tool for knowledge representation and reasoning in the Semantic Web. In this context, they are expected to play an important role in helping automated processes to access information. Namely: they are expected to provide a formal structure able to explicate the relationships between different concepts/terms, thus allowing intelligent agents to interpret, correctly, the semantics of the web resources improving the performances of the search technologies. Here we take into account a problem regarding Knowledge Representation in general, and ontology based representations in particular; namely: the fact that knowledge modeling seems to be constrained between conflicting requirements, such as compositionality, on the one hand and the need to represent prototypical information on the other. In particular, most common sense concepts seem not to be captured by the stringent semantics expressed by such formalisms as, for example, Description Logics (which are the formalisms on which the ontology languages have been built). The aim of this work is to analyse this problem, suggesting a possible solution suitable for formal ontologies and semantic web representations. The questions guiding this research, in fact, have been: is it possible to provide a formal representational framework which, for the same concept, combines both the classical modelling view (accounting for compositional information) and defeasible, prototypical knowledge ? Is it possible to propose a modelling architecture able to provide different type of reasoning (e.g. classical deductive reasoning for the compositional component and a non monotonic reasoning for the prototypical one)? We suggest a possible answer to these questions proposing a modelling framework able to represent, within the semantic web languages, a multilevel representation of conceptual information, integrating both classical and non classical (typicality based) information. Within this framework we hypothesise, at least in principle, the coexistence of multiple reasoning processes involving the different levels of representation.

As it emerged from philosophical analyses and cognitive research, most concepts exhibit typicality effects, and resist to the efforts of defining them in terms of necessary and sufficient conditions. This holds also in the case of many medical concepts. This is a problem for the design of computer science ontologies, since knowledge representation formalisms commonly adopted in this field do not allow for the representation of concepts in terms of typical traits. However, the need of representing concepts in terms of typical traits concerns almost every domain of real world knowledge, including medical domains. In particular, in this article we take into account the domain of mental disorders, starting from the DSM-5 descriptions of some specific mental disorders. On this respect, we favor a hybrid approach to the representation of psychiatric concepts, in which ontology oriented formalisms are combined to a geometric representation of knowledge based on conceptual spaces.

Combining typical knowledge to generate novel concepts is an important creative trait of human cognition. Dealing with such ability requires, from an AI perspective, the harmonization of two conflicting requirements that are hardly accommodated in symbolic systems: the need of a syntactic compositionality (typical of logical systems) and that one concerning the exhibition of typicality effects (see Frixione and Lieto, 2012). In this work we provide a logical framework able to account for this type of human-like concept combination. We propose a nonmonotonic Description Logic of typicality called TCL (Typicality-based Compositional Logic).

Endowing artificial systems with explanatory capacities about the reasons guiding their decisions, represents a crucial challenge and research objective in the current fields of Artificial Intelligence (AI) and Computational Cognitive Science [Langley et al., 2017]. Current mainstream AI systems, in fact, despite the enormous progresses reached in specific tasks, mostly fail to provide a transparent account of the reasons determining their behavior (both in cases of a successful or unsuccessful output). This is due to the fact that the classical problem of opacity in artificial neural networks (ANNs) explodes with the adoption of current Deep Learning techniques [LeCun, Bengio, Hinton, 2015]. In this paper we argue that the explanatory deficit of such techniques represents an important problem, that limits their adoption in the cognitive modelling and computational cognitive science arena. In particular we will show how the current attempts of providing explanations of the deep nets behaviour (see e.g. [Ritter et al. 2017] are not satisfactory. As a possibile way out to this problem, we present two different research strategies. The first strategy aims at dealing with the opacity problem by providing a more abstract interpretation of neural mechanisms and representations. This approach is adopted, for example, by the biologically inspired SPAUN architecture [Eliasmith et al., 2012] and by other proposals suggesting, for example, the interpretation of neural networks in terms of the Conceptual Spaces framework [Gärdenfors 2000, Lieto, Chella and Frixione, 2017]. All such proposals presuppose that the neural level of representation can be considered somehow irrelevant for attacking the problem of explanation [Lieto, Lebiere and Oltramari, 2017]. In our opinion, pursuing this research direction can still preserve the use of deep learning techniques in artificial cognitive models provided that novel and additional results in terms of “transparency” are obtained. The second strategy is somehow at odds with respect to the previous one and tries to address the explanatory issue by avoiding to directly solve the “opacity” problem. In this case, the idea is that one of resorting to pre-compiled plausible explanatory models of the word used in combination with deep-nets (see e.g. [Augello et al. 2017]). We argue that this research agenda, even if does not directly fits the explanatory needs of Computational Cognitive Science, can still be useful to provide results in the area of applied AI aiming at shedding light on the models of interaction between low level and high level tasks (e.g. between perceptual categorization and explanantion) in artificial systems.

In his famous 1982 paper, Allen Newell [22, 23] introduced the notion of knowledge level to indicate a level of analysis, and prediction, of the rational behavior of a cognitive articial agent. This analysis concerns the investigation about the availability of the agent knowledge, in order to pursue its own goals, and is based on the so-called Rationality Principle (an assumption according to which "an agent will use the knowledge it has of its environment to achieve its goals" [22, p. 17]. By using the Newell's own words: "To treat a system at the knowledge level is to treat it as having some knowledge, some goals, and believing it will do whatever is within its power to attain its goals, in so far as its knowledge indicates" [22, p. 13]. In the last decades, the importance of the knowledge level has been historically and system- atically downsized by the research area in cognitive architectures (CAs), whose interests have been mainly focused on the analysis and the development of mechanisms and the processes governing human and (articial) cognition. The knowledge level in CAs, however, represents a crucial level of analysis for the development of such articial general systems and therefore deserves greater research attention [17]. In the following, we will discuss areas of broad agree- ment and outline the main problematic aspects that should be faced within a Common Model of Cognition [12]. Such aspects, departing from an analysis at the knowledge level, also clearly impact both lower (e.g. representational) and higher (e.g. social) levels.

We propose a nonmonotonic Description Logic of typicality able to account for the phenomenon of combining prototypical concepts, an open problem in the fields of AI and cognitive modelling. Our logic extends the logic of typicality ALC + TR, based on the notion of rational closure, by inclusions p :: T(C) v D (“we have probability p that typical Cs are Ds”), coming from the distributed semantics of probabilistic Description Logics. Additionally, it embeds a set of cognitive heuristics for concept combination. We show that the complexity of reasoning in our logic is EXPTIME-complete as in ALC.

The paper introduces an extension of the proposal according to which conceptual representations in cognitive agents should be intended as heterogeneous proxytypes. The main contribution of this paper is in that it details how to reconcile, under a heterogeneous representational perspective, different theories of typicality about conceptual representation and reasoning. In particular, it provides a novel theoretical hypothesis - as well as a novel categorization algorithm called DELTA - showing how to integrate the representational and reasoning assumptions of the theory-theory of concepts with the those ascribed to the prototype and exemplars-based theories.