Tillmann Vierkant

This chapter discusses the question of whether there could ever be artificial intelligence with free will. This question reduces to whether or not artificial intelligence will ever be able to act on domain-general rationally formed intentions. However, it might not be possible to distinguish between behavior that is produced by such intentions and behavior that is merely a simulation. This could be because, as the Chinese Room thought experiment seems to show, consciousness intuitively is a necessary condition for intentionality. There are three potential responses to this challenge. One could argue that full domain-general rational behavior is possible only with consciousness, or that it might also be necessary for the system to be implemented in the right substrate, or that the right microfunctionalist structure needs to be in place for real intentionality.

The question of whether free will exists is often reduced to the binary question of whether or not its existence would be compatible with determinism. However, there are a number of important dimensions that permit different degrees of free will. Four components of free will seem to allow for a graded approach: (1) the ability to act in line with intentions (2) which are consciously formed (3) using one’s rational machinery and (4) without external interference. Finally, the chapter discusses whether there are lower and upper limits to this gradability such that above a certain level having more of the components does not create more free will, and below some threshold there is no free will at all, not just less.

In their intriguing paper Responsibility and Accountability in an Algorithmic Society (2025) the authors argue that the debate on how to deal with responsibility related issues with algorithmic agents requires a distinction between responsibility and accountability. In this comment to their paper, it is argued that while the notion of accountability as understood by the authors brings some significant benefits it also is ambiguous in an important way. Accountability could be understood as being purely instrumental with regard to general morally desirable consequences or it could be understood as necessarily containing an element of scaffolding for the agent who is held to account. The comment develops the options and discusses the consequences of choosing either of them.

Recent developments in the literature on mental agency have suggested that there are two quite different forms of mental agency. So-called managerial control can be understood simply along the lines of bodily intentional action. Evaluative control, on the other hand, is specific of mental agency. According to the proponents of this distinction, evaluative control is the primary or essential form of mental agency. In this chapter, it is argued that managerial control is more important than is assumed in the literature because it is essential for free mental agency. This, in turn, is a consequence of the fact that only managerial control allows for future-directed self-control properly understood. In order to justify this claim, the chapter then examines more closely the concept of managerial control and argues that we need to be very careful to not confound to similar-looking but very different capacities when defining managerial control.

The responsibility gap, commonly described as a core challenge for the effective governance of, and trust in, AI and autonomous systems (AI/AS), is traditionally associated with a failure of the epistemic and/or the control condition of moral responsibility: the ability to know what we are doing and exercise competent control over this doing. Yet these two conditions are a red herring when it comes to understanding the responsibility challenges presented by AI/AS, since evidence from the cognitive sciences shows that individual humans face very similar responsibility challenges with regard to these two conditions. While the problems of epistemic opacity and attenuated behaviour control are not unique to AI/AS technologies (though they can be exacerbated by them), we show that we can learn important lessons for AI/AS development and governance from how philosophers have recently revised the traditional concept of moral responsibility in response to these challenges to responsible human agency from the cognitive sciences. The resulting instrumentalist views of responsibility, which emphasize the forward-looking and flexible role of agency cultivation, hold considerable promise for integrating AI/AS into a healthy moral ecology. We note that there nevertheless is a gap in AI/AS responsibility that has yet to be extensively studied and addressed, one grounded in a relational asymmetry of vulnerability between human agents and sociotechnical systems like AI/AS. In the conclusion of this paper we note that attention to this vulnerability gap must inform and enable future attempts to construct trustworthy AI/AS systems and preserve the conditions for responsible human agency.

Orthodoxy has it that language evolution requires Gricean communicative intentions and therefore an understanding of nested metarepresentations. The problem with this orthodoxy is that it is hard to see how non-linguistic creatures could have such a sophisticated understanding of mentality. Some philosophers like Bar-On (The Journal of Philosophy 110 (6): 293-330, 2013a; Mind and Language 28 (3): 342-375, 2013b) have attempted to develop a non-Gricean account of language acquisition building on the information-rich and subtle communicative powers of expressive behaviours. This paper argues that this approach shares a number of instructive similarities with the literature on the cognitive conception of language. It shows how expressive accounts (and others) can learn from insights in that tradition and identifies a problem with the role of intentional actions in expressive accounts that can be solved by understanding expressive behaviours more as cognitive rather than communicative tools.

Epistemic agency is a crucial concept in many different areas of philosophy and the cognitive sciences. It is crucial in dual process theories of cognition as well as theories of metacognition and mindreading, self-control, and moral agency. But what is epistemic agency? The Tinkering Mind argues that epistemic agency has two distinct and incompatible definitions. It can be simply understood as intentional mental action, or as a distinct non-voluntary form of evaluative agency. The core argument of the book demonstrates that both definitions lead to surprising and counterintuitive consequences. If epistemic agency is a form of intentional action, then this implies that the radical theory of extended cognition has to be true. If, on the other hand, epistemic agency is not intentional action but evaluative agency, then intentional epistemic actions like deliberation are not truly cognitive but merely catalytic. Once established, the distinction between these two options sheds new light on various and diverse philosophical and psychological debates from dual process theories to debates on choice and self-control.

A long-standing position in philosophy, law, and theology is that a person can be held morally responsible for an action only if they had the freedom to choose and to act otherwise. Thus, many philosophers consider freedom to be a necessary condition for moral responsibility. However, empirical findings suggest that this assumption might not be in line with common sense thinking. For example, in a recent study we used surveys to show that – counter to positions held by many philosophers – lay people consider actions to be free when they are spontaneous rather than being based on reasons. In contrast, responsibility is often considered to require that someone has thought about the alternative options. In this study we use an online survey to directly test to which degree lay judgements of freedom and responsibility match. Specifically, we test whether manipulations of deliberation affect freedom and responsibility judgements in the same way. Furthermore, we also test the dependency of these judgements on a person’s belief that their decision had consequences for their personal life. We find that people are generally considered to be more responsible for their actions than they are considered free, across most of the conditions tested. Furthermore, we found that deliberation had an opposite effect on freedom and responsibility judgements. People were considered more free when they acted spontaneously, whereas they were considered more responsible when they deliberated about their actions. These results seem to suggest that deliberating about reasons is crucially important for the lay concept of responsibility, while it is not obviously …

The article argues that there is a specific role for narrative consciousness in our understanding of justified responsibility ascription. Starting from a short review of empirical findings that suggest that we do not consciously control our actions, the article proceeds to spell out a concept of willed actions that does justice to the scientific results, conceptual requirements, and our most important intuitions on the ascription of responsibility. In order to do this, the article develops a concept of how narrative monitoring can be said to control actions that are performed on very small time scales. This concept understands narrative monitoring as the self interpretation and disambiguation of complex cognitive systems. Finally, consequences of this concept for the ascription of responsibility are explored.

Intuitively, choices seem to be intentional actions but it is difficult to see how they could be. If our choices are all about weighing up reasons then there seems no room for an additional intentional act of choice. Richard Holton has suggested a solution to this puzzle, which involves thinking of choices in a two systems of cognition framework. Holton’s suggestion does solve the puzzle, but has some unsatisfactory consequences. This paper wants to take over the important insights from Holton on the phenomenology of choice and the possible explanation via a two systems framework, but wants to suggest an alternative more satisfactory account. This account is built around the idea that choices are what Pamela Hieronymi calls managerial acts. After developing the claim the paper then defends it against the objection that managerial control cannot be understood in a system1 context, by examining recent research on uncertainty monitoring and early forms of metacognition.

This book gives a convincing philosophical explanation for the strong persistence of our diverse folk psychological intuitions about the self. For this purpose it introduces, on the one hand, the distinction between subject and self model as proposed by Metzinger, on the other hand, the distinction between a social/normative and a cognitive/organic perspective on the self. The book argues that one needs to take into account both distinctions, if one wants to answer notoriously difficult questions like the one that gives the book its title: Is the self real?

This paper is trying to show that it is not possible to use the Knowledge argument as independent evidence for the form of non-reductionism the Modal argument argues for. To show this, Jackson's famous 'Mary' thought experiment is imagined in a zombie world. This leads to the result that there are many problems in the Mary experiment, which cannot have anything to do with phenomenal Qualia, because the Zombie-Mary would encounter them as well, and once all these problems are accounted for, it is no longer clear whether a Zombie-Mary is conceivable at all. Finally, an alternative explanation for the strong non-reductive intuitions of the Mary experiment is discussed

Understanding and controlling our minds is one of the most fascinating features of human cognition. It has often been assumed that this ability requires a theoretical understanding of psychological states. This assumption has recently been put under pressure by so called mindshaping approaches. We agree that these approaches provide us with a new way of self-understanding and that they enable a very powerful form of self-regulation which we label narrative control. However, we insist that there still is a crucial role for a theoretical understanding of psychological states in human cognition. We argue that this is because a theoretical understanding is necessary for all intentional control of the mind. Keywords: mindreading; mindshaping; narrative control; manipulative control; folk psychology

This paper argues against the intuition that willpower and so called ‘tying to the mast’ strategies are fundamentally different types of mental actions to achieve self control. The argument for this surprising claim is that at least on the most plausible account of willpower an act of willpower consists in an intentional mental action that disables the mental agent and thereby creates a mental tie. The paper then defends this claim against the objection that tying to the mast strategies do not have the same phenomenology as real willpower and that they do not preserve reason responsiveness

Holding content explicitly requires a form of self knowledge. But what does the relevant self knowledge look like? Using theory of mind as an example, this paper argues that the correct answer to this question will have to take into account the crucial role of language based deliberation, but warns against the standard assumption that explicitness is necessary for ascribing awareness. It argues in line with Bayne that intentional action is at least an equally valid criterion for awareness. This leads to a distinction between different levels of implicitness. Postulating these different levels, it is argued, allows us to make better sense of the empirical literature on early false belief task abilities.

Research in psychology clearly shows that implicit biases contribute significantly to our behaviour. What is less clear, however, is whether we are responsible for our implicit biases in the same way that we are responsible for our explicit beliefs. Neil Levy has argued recently that explicit beliefs are special with regard to the responsibility we have for them, because they unify the agent. In this paper we point out multiple ways in which implicit biases also unify the agent. We then examine Levy’s claim that the assertibility of explicit beliefs means that they have a unique way of unifying the agent by being available for syntactical operations. We accept that syntactical operations are important, but worry that they are less straightforwardly connected to the unification of agents than Levy claims

It is commonly assumed in the philosophical literature that in order to acquire an intention, the agent has to settle a question of what to do in practical deliberation. Carruthers, P. has recently used this to argue that the acquisition of intentions can never be conscious even in cases where the agent asserts having the intention in inner speech. Because of that Carruthers also believes that knowledge of intentions even in first person cases is observational. This paper explores the challenge Carruthers’ argument throws up for accounts that also rely on the notion of settling a question for intention acquisition, but who also want to maintain at the same time that knowledge of intentions in the first person case is not observational.

In the wake of Clark and Chalmers famous argument for extended cognition some people have argued that willpower equally can extend into the environment (e.g. Heath and Anderson in The thief of time: philosophical essays on procrastination. Oxford University Press, New York, pp 233–252, 2010). In a recent paper Fabio Paglieri (Consciousness in interaction: the role of the natural and social context in shaping consciousness. John Benjamins, Amsterdam, pp 179–206, 2012) provides an interesting argument to the effect that there might well be extended self control, but that willpower does not lend itself to extension. This paper argues that Paglieri is right in claiming that previous attempts to extend the will are flawed. It then provides an argument for extending the will that does not fall foul of Paglieri’s argument and actually provides us with an even stronger case for extension than the one that Clark and Chalmers provide for cognition

Humans seem special, because they are able to think about thinking (to make their mentality the object of their thoughts). In this paper I distinguish two very different ways in which thinking about thinking could be understood and which role these understandings play in the control of the mind. I argue on the one hand that language enables humans to express and attend to the content of their thoughts. This is an incredibly powerful tool which allows for the intentional manipulation of content. On the other hand, however, this form of thinking about thinking does not require the agent to understand mental states as states and because of that it does not allow for the intentional targeting of the states. Only a second and metarepresentation based form of thinking about thinking allows for this control. the rest of the paper then discusses the possible function of the intentional control of the mind in future directed self control.