Dimitri Coelho Mollo

Artificial Intelligence is a field that lives many lives, and the term has come to encompass a motley collection of scientific and commercial endeavours. In this paper, I articulate the contours of a rather neglected but central scientific role that AI has to play, which I dub “AI-as-exploration”. The basic thrust of AI-as-explora tion is that of creating and studying systems that can reveal candidate building blocks of intelligence that may dif fer from the kinds of human and animal intelligence we are familiar with. In other words, I suggest that AI is one of the best tools we have for exploring intelligence space, namely the space of possible intelligent systems. I illus trate the value of AI-as-exploration by focusing on a specific case study, i. e., recent work on the capacity to com bine novel and invented concepts in humans and Large Language Models. I show that the latter, despite showing human-level accuracy in such a task, probably solve it in ways radically different, but no less relevant to intelli gence research, to those hypothesised for humans.

Large language models (LLMs) produce seemingly meaningful outputs, yet they are trained on text alone without direct interaction with the world. This leads to a modern variant of the classical symbol grounding problem in AI: can LLMs' internal states and outputs be about extra-linguistic reality, independently of the meaning human interpreters project onto them? We argue that they can. We first distinguish referential grounding—the connection between a representation and its worldly referent—from other forms of grounding and argue it is the only kind essential to solving the problem. We contend that referential grounding is achieved when a system's internal states satisfy two conditions derived from teleosemantic theories of representation: (1) they stand in appropriate causal-informational relations to the world, and (2) they have a history of selection that has endowed them with the function of carrying this information. We argue that LLMs can meet both conditions, even without multimodality or embodiment.

This article investigates how human interactions with AI‐powered chatbots may offend human dignity. Current chatbots, driven by large language models, mimic human linguistic behaviour but lack the moral and rational capacities essential for genuine interpersonal respect. Human beings are prone to anthropomorphize chatbots – indeed, chatbots appear to be deliberately designed to elicit that response. As a result, human beings' behaviour towards chatbots often resembles behaviours typical of interaction between moral agents. Drawing on a second‐personal, relational account of dignity, we argue that interacting with chatbots in this way is incompatible with the dignity of users. We show that, since second‐personal respect is premised on reciprocal recognition of second‐personal moral authority, behaving towards chatbots in ways that convey second‐personal respect is bound to misfire in morally problematic ways, given the lack of reciprocity. Consequently, such chatbot interactions amount to subtle but significant violations of self‐respect – the respect we are duty‐bound to show for our own dignity. We illustrate this by discussing four actual chatbot use cases (information retrieval, customer service, advising, and companionship), and propound that the increasing societal pressure to engage in such interactions with chatbots poses a hitherto underappreciated threat to human dignity.

Crane (2024) argues against what he calls the ‘aboutness assumption’, which he takes to be the basic position in contemporary philosophy of mind. According to this assumption, in order to solve the problem of aboutness we must find a single account that explains how the various types of representation (external and internal) come to be about states and events in the world. For Crane, the aboutness assumption is underpinned by the idea that there is an essential similarity between external and internal representations when it comes to their aboutness—based, in its turn, on strong, debatable commitments, for example, to physicalism about the mental.I present two lines of criticism of Crane’s analysis. First, I show that many influential philosophical theories of mind and cognition do not subscribe to the aboutness assumption. Second, I argue that the idea that there is an essential similarity between external and internal representations does not hinge on strong philosophical pre-commitments, but rather on the extension of a successful model of explanation of complex behaviour based on external representations to the mental, internal case. This realisation, moreover, opens up considerable space for the kind of pluralism about intentionality that Crane advocates.

In the literature on the nature and role of cognitive representation, three positions are taken across the conceptual landscape: robust realism, primitivism, and eliminativism. Recently, a fourth alternative that tries to avoid the shortcomings of traditional views has been proposed: content pragmatism. My aim is to defend pragmatism about content against some recent objections moved against the view. According to these objections, content pragmatism (a) fails to capture the role played by representation in the cognitive sciences; and/or (b) is an unstable view that ends up collapsing into one of the traditional alternatives. I argue that those arguments fail. I show that content pragmatism has as much claim to descriptive adequacy as the traditional theories. Moreover, I defend the robustness of the view by arguing that it does not collapse into any of the traditional positions. Content pragmatism therefore offers a valid and coherent account of the nature of representational content.

The remarkable performance of large language models (LLMs) on complex linguistic tasks has sparked debate about their capabilities. Unlike humans, these models learn language solely from textual data without directly interacting with the world. Yet they generate seemingly meaningful text on diverse topics. This achievement has renewed interest in the classical `Symbol Grounding Problem' -- the question of whether the internal representations and outputs of symbolic AI systems can possess intrinsic meaning that is not parasitic on external interpretation. Although modern LLMs compute over vectors rather than symbols, an analogous problem arises for these systems, which we call the Vector Grounding Problem. This paper has two main goals. First, we distinguish five main notions of grounding that are often conflated in the literature, and argue that only one of them, which we call referential grounding, is relevant to the Vector Grounding Problem. Second, drawing on philosophical theories of representational content, we provide two arguments for the claim that LLMs and related systems can achieve referential grounding: (1) through preference fine-tuning methods that explicitly establish world-involving functions, and (2) through pre-training alone, which in limited domains may select for internal states with world-involving content, as mechanistic interpretability research suggests. Through these pathways, LLMs can establish connections to the world sufficient for intrinsic meaning. One potentially surprising implication of our discussion is that that multimodality and embodiment are neither necessary nor sufficient to overcome the Grounding Problem. *Note: this is an updated and improved version of the preprint originally made available on ArXiv in 2023.*

This paper critically evaluates the attempts to align Artificial Intelligence (AI) systems, especially Large Language Models (LLMs), with human values and intentions through Reinforcement Learning from Feedback methods, involving either human feedback (RLHF) or AI feedback (RLAIF). Specifically, we show the shortcomings of the broadly pursued alignment goals of honesty, harmlessness, and helpfulness. Through a multidisciplinary sociotechnical critique, we examine both the theoretical underpinnings and practical implementations of RLHF techniques, revealing significant limitations in their approach to capturing the complexities of human ethics, and contributing to AI safety. We highlight tensions inherent in the goals of RLHF, as captured in the HHH principle (helpful, harmless and honest). In addition, we discuss ethically-relevant issues that tend to be neglected in discussions about alignment and RLHF, among which the trade-offs between user-friendliness and deception, flexibility and interpretability, and system safety. We offer an alternative vision for AI safety and ethics which positions RLHF approaches within a broader context of comprehensive design across institutions, processes and technological systems, and suggest the establishment of AI safety as a sociotechnical discipline that is open to the normative and political dimensions of artificial intelligence.

Artificial Intelligence is a field that lives many lives, and the term has come to encompass a motley collection of scientific and commercial endeavours. In this paper, I articulate the contours of a rather neglected but central scientific role that AI has to play, which I dub “AI-as-exploration”. The basic thrust of AI-as-exploration is that of creating and studying systems that can reveal candidate building blocks of intelligence that may differ from the forms of human and animal intelligence we are familiar with. In other words, I suggest that AI is one of the best tools we have for exploring intelligence space, namely the space of possible intelligent systems. I illustrate the value of AI-as-exploration by focusing on a specific case study, i.e., recent work on the capacity to combine novel and invented concepts in humans and Large Language Models. I show that the latter, despite showing human-level accuracy in such a task, most probably solve it in ways radically different, but no less relevant to intelligence research, to those hypothesised for humans.

This article investigates how human interactions with AI-powered chatbots may offend human dignity. Current chatbots, driven by large language models, mimic human linguistic behaviour but lack the moral and rational capacities essential for genuine interpersonal respect. Human beings are prone to anthropomorphize chatbots – indeed, chatbots appear to be deliberately designed to elicit that response. As a result, human beings' behaviour towards chatbots often resembles behaviours typical of interaction between moral agents. Drawing on a second-personal, relational account of dignity, we argue that interacting with chatbots in this way is incompatible with the dignity of users. We show that, since second-personal respect is premised on reciprocal recognition of second-personal moral authority, behaving towards chatbots in ways that convey second-personal respect is bound to misfire in morally problematic ways, given the lack of reciprocity. Consequently, such chatbot interactions amount to subtle but significant violations of self-respect – the respect we are duty-bound to show for our own dignity. We illustrate this by discussing four actual chatbot use cases (information retrieval, customer service, advising, and companionship), and propound that the increasing societal pressure to engage in such interactions with chatbots poses a hitherto underappreciated threat to human dignity.

Abstract: Research on the nature and varieties of the format of cognitive representations in philosophy and cognitive science have been partly shaped by analogies to external, public representations. In this paper, we argue that relying on such analogies contributes to framing the question of cognitive formats in problematic, potentially counterproductive ways. We show that cognitive and public representations differ in many of their central features, making analogies to public representations ill-suited to improving our understanding of cognitive formats. We illustrate these points by examining two case studies in which analogies to public representations may have had a negative impact on research: the 80’s-90’s debate about compositionality and cognitive architecture between symbolicists and connectionists; and contemporary discussions about the nature of visual demonstratives. Finally, we outline an alternative, computational account of formats that does not share the shortcomings of appeal to public representations.

The notion of intelligence is relevant to several fields of research, including cognitive and comparative psychology, neuroscience, artificial intelligence, and philosophy, among others. However, there is little agreement within and across these fields on how to characterise and explain intelligence. I put forward a behavioural, operational characterisation of intelligence that can play an integrative role in the sciences of intelligence, as well as preserve the distinctive explanatory value of the notion, setting it apart from the related concepts of cognition and rationality. Finally, I examine a popular hypothesis about the underpinnings of intelligence: the capacity to manipulate internal representations of the environment. I argue that the hypothesis needs refinement, and that so refined, it applies only to some forms of intelligence.

Debate on the nature of representation in cognitive systems tends to oscillate between robustly realist views and various anti‐realist options. I defend an alternative view, deflationary realism, which sees cognitive representation as an offshoot of the extended application to cognitive systems of an explanatory model whose primary domain is public representation use. This extended application, justified by a common explanatory target, embodies idealisations, partial mismatches between model and reality. By seeing representation as part of an idealised model, deflationary realism avoids the problems with robust realist views, while keeping allegiance to realism.

Cognitive representations are typically analysed in terms of content, vehicle and format. While current work on formats appeals to intuitions about external representations, such as words and maps, in this paper we develop a computational view of formats that does not rely on intuitions. In our view, formats are individuated by the computational profiles of vehicles, i.e., the set of constraints that fix the computational transformations vehicles can undergo. The resulting picture is strongly pluralistic, it makes space for a variety of different formats, and is intimately tied to the computational approach to cognition in cognitive science and artificial intelligence.

Fairness is central to the ethical and responsible development and use of AI systems, with a large number of frameworks and formal notions of algorithmic fairness being available. However, many of the fairness solutions proposed revolve around technical considerations and not the needs of and consequences for the most impacted communities. We therefore want to take the focus away from definitions and allow for the inclusion of societal and relational aspects to represent how the effects of AI systems impact and are experienced by individuals and social groups. In this paper, we do this by means of proposing the ACROCPoLis framework to represent allocation processes with a modeling emphasis on fairness aspects. The framework provides a shared vocabulary in which the factors relevant to fairness assessments for different situations and procedures are made explicit, as well as their interrelations. This enables us to compare analogous situations, to highlight the differences in dissimilar situations, and to capture differing interpretations of the same situation by different stakeholders.

An influential view in cognitive science is that computation in cognitive systems is semantic, conceptually depending on representation: to compute is to manipulate representations. I argue that accepting the non-semantic teleomechanistic view of computation lays the ground for a promising alternative strategy, in which computation helps to explain and naturalise representation, rather than the other way around. I show that this computation-based approach to representation presents six decisive advantages over the semantic view. I claim that it can improve the two most influential current theories of representation, teleosemantics and structural representation, by providing them with precious tools to tackle some of their main shortcomings. In addition, the computation-based approach opens up interesting new theoretical paths for the project of naturalising representation, in which teleology plays a role in individuating computations, but not representations.

Computational perspectivalism has been recently proposed as an alternative to mainstream accounts of physical computation, and especially to the teleologically-based mechanistic view. It takes physical computation to be partly dependent on explanatory perspectives and eschews appeal to teleology in helping individuate computational systems. I assess several varieties of computational perspectivalism, showing that they either collapse into existing non-perspectival views or end up with unsatisfactory or implausible accounts of physical computation. Computational perspectivalism fails, therefore, to be a compelling alternative to perspective-independent theories of computation in physical systems. I conclude that a teleologically-based, non-perspectival mechanistic account of physical computation is to be preferred. 1Introduction 2The Mechanistic View of Computation 2.1Teleological functions and meeting the desiderata 3Varieties of Perspectivalism 4Computational Perspectivalism 4.1Computational perspectivalism and pancomputationalism 4.2Computational perspectivalism and miscomputation 5The Varieties of Computational Perspectivalism Assessed 5.1Innocuous computational perspectivalism 5.2Non-innocuous computational perspectivalism 5.2.1Computational instrumentalism 5.2.2Ontic perspectivalism 6Concluding Remarks

I analyze a tension at the core of the mechanistic view of computation generated by its joint commitment to the medium independence of computational vehicles and to computational systems possessing teleological functions to compute. While computation is individuated in medium-independent terms, teleology is sensitive to the constitutive physical properties of vehicles. This tension spells trouble for the mechanistic view, suggesting that there can be no teleological functions to compute. I argue that, once considerations about the relevant function-bestowing factors for computational systems are brought to bear, the tension dissolves: physical systems can have the teleological function to compute.

In the literature on the nature and role of cognitive representation, three positions are taken across the conceptual landscape: robust realism, primitivism, and eliminativism. Recently, a fourth alternative that tries to avoid the shortcomings of traditional views has been proposed: content pragmatism. My aim is to defend pragmatism about content against some recent objections moved against the view. According to these objections, content pragmatism fails to capture the role played by representation in the cognitive sciences; and/or is an unstable view that ends up collapsing into one of the traditional alternatives. I argue that those arguments fail. I show that content pragmatism has as much claim to descriptive adequacy as the traditional theories. Moreover, I defend the robustness of the view by arguing that it does not collapse into any of the traditional positions. Content pragmatism therefore offers a valid and coherent account of the nature of representational content.

In the target article Hutto and Satne propose a new approach to studying mental content. Although I believe there is much to commend in their proposal, I argue that it makes no space for a kind of content that is of central importance to cognitive science, and which need not be involved in beliefs and desires: I will use the expression ‘representational content’ to refer to it. Neglecting representational content leads to an undue limitation of the contribution that the neo-Cartesian approach can offer to the naturalising content project. I claim that neo-Cartesians can, on the one hand, help account for the nature of representational content and clarify what makes representational states contentful. On the other, besides explaining the natural origins of Ur-intentionality, neo-Cartesians should also take the role of accounting for the natural origins of contentful states that fall short of beliefs and desires. Finally, I argue that the only alternative for the authors is to embrace some form of non-representationalism, as Hutto elsewhere does. The success of the proposal thereby turns on the fate of the radical non-representationalist position that it accompanies

I examine a major objection to the mechanistic view of concrete computation, stemming from an apparent tension between the abstract nature of computational explanation and the tenets of the mechanistic framework: while computational explanation is medium-independent, the mechanistic framework insists on the importance of providing some degree of structural detail about the systems target of the explanation. I show that a common reply to the objection, i.e. that mechanistic explanation of computational systems involves only weak structural constraints, is not enough to save the standard mechanistic view of computation—it trivialises the appeal to mechanism, and thus makes the account collapse into a purely functional view. I claim, however, that the objection can be put to rest once the account is appropriately amended: computational individuation is indeed functional, while mechanistic explanation plays a role in accounting for computational implementation. Since individuation and implementation are crucial elements in a satisfying account of computation in physical systems, mechanism keeps its central importance in the theory of concrete computation. Finally, I argue that my version of the mechanistic view helps to provide a convincing reply to a powerful objection against non-semantic theories of concrete computation: the argument from the multiplicity of computations.