In this opinionated review, I draw attention to some of the contributions reinforcement learning can make to questions in the philosophy of mind. In particular, I highlight reinforcement learning's foundational emphasis on the role of reward in agent learning, and canvass two ways in which the framework may advance our understanding of perception and motivation.

In their paper 'Reward is enough', Silver et al conjecture that the creation of sufficiently good reinforcement learning (RL) agents is a path to artificial general intelligence (AGI). We consider one aspect of intelligence Silver et al did not consider in their paper, namely, that aspect of intelligence involved in designing RL agents. If that is within human reach, then it should also be within AGI's reach. This raises the question: is there an RL environment which incentivises RL (...) agents to design RL agents? (shrink)

After generalizing the Archimedean property of real numbers in such a way as to make it adaptable to non-numeric structures, we demonstrate that the real numbers cannot be used to accurately measure non-Archimedean structures. We argue that, since an agent with Artificial General Intelligence (AGI) should have no problem engaging in tasks that inherently involve non-Archimedean rewards, and since traditional reinforcement learning rewards are real numbers, therefore traditional reinforcement learning probably will not lead to AGI. We (...) indicate two possible ways traditional reinforcement learning could be altered to remove this roadblock. (shrink)

We consider an extended notion of reinforcement learning in which the environment can simulate the agent and base its outputs on the agent's hypothetical behavior. Since good performance usually requires paying attention to whatever things the environment's outputs are based on, we argue that for an agent to achieve on-average good performance across many such extended environments, it is necessary for the agent to self-reflect. Thus weighted-average performance over the space of all suitably well-behaved extended environments could be (...) considered a way of measuring how self-reflective an agent is. We give examples of extended environments and introduce a simple transformation which experimentally seems to increase some standard RL agents' performance in a certain type of extended environment. (shrink)

Background Men with antisocial personality disorder show lifelong abnormalities in adaptive decision making guided by the weighing up of reward and punishment information. Among men with antisocial personality disorder, modifi cation of the behaviour of those with additional diagnoses of psychopathy seems particularly resistant to punishment. Methods We did a case-control functional MRI (fMRI) study in 50 men, of whom 12 were violent off enders with antisocial personality disorder and psychopathy, 20 were violent off enders with antisocial personality disorder but (...) not psychopathy, and 18 were healthy non-off enders. We used fMRI to measure brain activation associated with the representation of punishment or reward information during an event-related probabilistic response-reversal task, assessed with standard general linear-model-based analysis. Findings Offenders with antisocial personality disorder and psychopathy displayed discrete regions of increased activation in the posterior cingulate cortex and anterior insula in response to punished errors during the task reversal phase, and decreased activation to all correct rewarded responses in the superior temporal cortex. This finding was in contrast to results for off enders without psychopathy and healthy non-off enders. Interpretation Punishment prediction error signalling in off enders with antisocial personality disorder and psychopathy was highly atypical. This finding challenges the widely held view that such men are simply characterised by diminished neural sensitivity to punishment. Instead, this fi nding indicates altered organisation of the information processing system responsible for reinforcement learning and appropriate decision making. This difference between violent offenders with antisocial personality disorder with and without psychopathy has implications for the causes of these disorders and for treatment approaches. (shrink)

This study aimed to identify the learning and business incubation processes and their impact on developing the performance of business incubators in Gaza Strip, and the study relied on the descriptive analytical approach, and the study population consisted of all employees working in business incubators in Gaza Strip in addition to experts and consultants in incubators where their total number reached (62) individuals, and the researchers used the questionnaire as a main tool to collect data through the comprehensive survey (...) method, where (55) questionnaires were retrieved with a recovery rate of (88.7%). The results of the study showed that there is a high approval of the variables of the study, where the dimension of incubation operations came with a relative weight (84.89%), followed by the learning dimension which got a relative weight (82.50%), while the performance level of the incubators got a relative weight (80.12%) and as the study results showed There is a positive relationship statistically between the two dimensions of learning and business incubation processes and performance development in business incubators, and the linear regression analysis model showed that the main effect in improving the performance of business incubators is "after business incubation operations" while it showed weak effect after learning. The study came out with several recommendations, the most important of which is that business incubators seek to reinforce the learning process and incubation processes for business, and the need to periodically evaluate their performance, and that a more effective mechanism be designed to follow up with companies after the end of the incubation period and monitor the progress of these companies. (shrink)

This paper presents a human-computer interaction model with a three layers learning mechanism in a pervasive environment. We begin with a discussion around a number of important issues related to human-computer interaction followed by a description of the architecture for a multi-agent cooperative design system for pervasive computing environment. We present our proposed three- layer HCI model and introduce the group formation algorithm, which is predicated on a dynamic sharing niche technology. Finally, we explore the cooperative reinforcement (...) class='Hi'>learning and fusion algorithms; the paper closes with concluding observations and a summary of the principal work and contributions of this paper. (shrink)

In modern, culturally-heterogeneous societies, inefficiency of communication of important moral concepts is often evidenced by asymmetrical moral judgements and hypocritical behaviour, especially in our increasingly compartmentalised social landscapes [Rozuel 2011]. This raises the question of how to present target audiences with some (perhaps novel) moral concept, like an ethical dimension of one’s ecological attitude, in a way which would resonate with them, and be conducive to a coherent moral stance, decreasing action-observer biases. We analyse this problem by introducing a formal (...) moral inconsistency framework. We consider the ways in which the aforementioned problem might be tackled in existing cognitive science of morality (CSM) nativist frameworks, and, secondly, we investigate it from the perspective of moral learning, finally proposing the notion of moral dimensionality of a representation in the context of model-based reinforcement learning, as a useful conceptual tool for studying the problem. (shrink)

A new trend in deep learning, represented by Mutual Information Neural Estimation (MINE) and Information Noise Contrast Estimation (InfoNCE), is emerging. In this trend, similarity functions and Estimated Mutual Information (EMI) are used as learning and objective functions. Coincidentally, EMI is essentially the same as Semantic Mutual Information (SeMI) proposed by the author 30 years ago. This paper first reviews the evolutionary histories of semantic information measures and learning functions. Then, it briefly introduces the author’s semantic information (...) G theory with the rate-fidelity function R(G) (G denotes SeMI, and R(G) extends R(D)) and its applications to multi-label learning, the maximum Mutual Information (MI) classification, and mixture models. Then it discusses how we should understand the relationship between SeMI and Shannon’s MI, two generalized entropies (fuzzy entropy and coverage entropy), Autoencoders, Gibbs distributions, and partition functions from the perspective of the R(G) function or the G theory. An important conclusion is that mixture models and Restricted Boltzmann Machines converge because SeMI is maximized, and Shannon’s MI is minimized, making information efficiency G/R close to 1. A potential opportunity is to simplify deep learning by using Gaussian channel mixture models for pre-training deep neural networks’ latent layers without considering gradients. It also discusses how the SeMI measure is used as the reward function (reflecting purposiveness) for reinforcement learning. The G theory helps interpret deep learning but is far from enough. Combining semantic information theory and deep learning will accelerate their development. (shrink)

Are the real numbers rich enough to measure intelligence? We generalize a result of Alexander and Hutter about the so-called Legg-Hutter intelligence measures of reinforcement learning agents. Using the generalized result, we exhibit a paradox: in one particular version of the Legg-Hutter intelligence measure, certain agents all have intelligence 0, even though in a certain sense some of them outperform others. We show that this paradox disappears if we vary the Legg-Hutter intelligence measure to be hyperreal-valued rather than (...) real-valued. (shrink)

We present a game mechanic called pseudo-visibility for games inhabited by non-player characters (NPCs) driven by reinforcement learning (RL). NPCs are incentivized to pretend they cannot see pseudo-visible players: the training environment simulates an NPC to determine how the NPC would act if the pseudo-visible player were invisible, and penalizes the NPC for acting differently. NPCs are thereby trained to selectively ignore pseudo-visible players, except when they judge that the reaction penalty is an acceptable tradeoff (e.g., a guard (...) might accept the penalty in order to protect a treasure because losing the treasure would hurt even more). We describe an RL agent transformation which allows RL agents that would not otherwise do so to perform some limited self-reflection to learn the training environments in question. (shrink)

The results are interpreted in terms of a deficit in initial rule learning and subsequent generalization of these rules to new stimuli. Negative feedback is adequately processed at a neural level but this information is not used to improve behaviour on subsequent trials. As learning is degraded, the process of error detection at the moment of the actual response is diminished. Therefore, the current study demonstrates that disturbed error-monitoring processes play a central role in the often reported (...) class='Hi'>learning deficits in individuals with PP. (shrink)

Curiosity is evident in humans of all sorts from early infancy, and it has also been said to appear in a wide range of other animals, including monkeys, birds, rats, and octopuses. The classical definition of curiosity as an intrinsic desire for knowledge may seem inapplicable to animal curiosity: one might wonder how and indeed whether a rat could have such a fancy desire. Even if rats must learn many things to survive, one might expect their learning must be (...) driven by simpler incentives, such as hunger. One might also wonder what proximal signals could guide animals towards knowledge itself, or how something as abstract as knowledge could ever be a motivational target for an unreflective animal. Taking a cue from recent work in reinforcement learning, I argue that surprise functions as a reward signal for the curious animal, and then show how this amounts to a desire for knowledge gain, where knowledge is conceived of as a cognitive adaptation to reality. This adaptation results in a mental state whose existence depends essentially on the truth of its contents, a factive mental state. Curious creatures benefit from an interaction between the prediction-error correction processes of basic learning and the active surprise-seeking force of their curiosity. This internally adversarial interaction accelerates knowledge gain in ways that are very helpful for agents with the restrictions of biological creatures, in environments with the complexity of our natural world. (shrink)

What is common knowledge? According to the dominant iterative model, a group of people commonly knows that p if and only if they each individually know that p, and they furthermore each know that they each know that p, and so on to infinity. According to the integrative model proposed in this paper, a group commonly knows that p when its members are united in a state of mind of the type whose contents must be true. Epistemic integration within a (...) group is enabled by symmetrical signalling processes such as eye contact. In conversational dyads, symmetrical processing operates on pairs of signals produced by the two sides in a familiar format: speakers generate content for joint attention in main channel communication, and addressees evaluate that content in backchannel communication. Processes of reinforcement learning shape our pairwise signalling, driving the accumulation of common knowledge, both in response to extrinsic reward for coordinated action, and in response to the intrinsic reward of curiosity. Where the iterative model caps the epistemic performance of the group at the level of its weakest member, the integrative model of common knowledge shows how groups working together can outperform their strongest member working alone. (shrink)

I claim that the significance of comic works to influence our attitudes is limited by the conditions under which we find things funny. I argue that we can only find something funny if we regard it as norm-violating in a way that doesn’t make certain cognitive or pragmatic demands upon us. It is compatible with these conditions that humour reinforces our attitude that something is norm-violating. However, it is not compatible with these conditions that, on the basis of finding it (...) funny, we come to reject some existing attitude. Such a rejection would require that we recognize our attitude as norm-violating in a way that has pragmatic force. Thus if a humorous work reveals the absurdity of something, we can either find it funny and not have our attitudes significantly influenced, or else be significantly influenced but not find it funny. (shrink)

I propose that the successes and contributions of reinforcement learning urge us to see the mind in a new light, namely, to recognise that the mind is fundamentally evaluative in nature.

Mathematicians and software developers use the word "function" very differently, and yet, sometimes, things that are in practice implemented using the software developer's "function", are mathematically formalized using the mathematician's "function". This mismatch can lead to inaccurate formalisms. We consider a special case of this meta-problem. Various kinds of agents might, in actual practice, make use of private memory, reading and writing to a memory-bank invisible to the ambient environment. In some sense, we humans do this when we silently subvocalize (...) thoughts about the actions we are taking (at least when the environment we're in is too primitive to probe the contents of our brains). Mathematical function formalizations of agents often ignore this ability. We show that in a general agent-environment framework (of which reinforcement learning is a special case), in a technical sense, such private memories do not enable qualitatively different agent behavior. (shrink)

Inspired by recent progress in multi-agent Reinforcement Learning (RL), in this work we examine the collective intelligent behaviour of theoretical universal agents by introducing a weighted mixture operation. Given a weighted set of agents, their weighted mixture is a new agent whose expected total reward in any environment is the corresponding weighted average of the original agents' expected total rewards in that environment. Thus, if RL agent intelligence is quantified in terms of performance across environments, the weighted mixture's (...) intelligence is the weighted average of the original agents' intelligences. This operation enables various interesting new theorems that shed light on the geometry of RL agent intelligence, namely: results about symmetries, convex agent-sets, and local extrema. We also show that any RL agent intelligence measure based on average performance across environments, subject to certain weak technical conditions, is identical (up to a constant factor) to performance within a single environment dependent on said intelligence measure. (shrink)

In this paper, I examine Reinforcement Learning modelling practice in psychiatry, in the context of alcohol use disorders. I argue that the epistemic roles RL currently plays in the development of psychiatric classification and search for explanations of clinically relevant phenomena are best appreciated in terms of Chang’s account of epistemic iteration, and by distinguishing mechanistic and aetiological modes of computational explanation.

Where do values live in thought? A straightforward answer is that we (or our brains) make decisions using explicit value representations which are our values. Recent work applying reinforcement learning to decision-making and planning suggests that more specifically, we may represent both the instrumental expected value of actions as well as the intrinsic reward of outcomes. In this paper, I argue that identifying value with either of these representations is incomplete. For agents such as humans and other animals, (...) there is another place where reward can be located in thought: the division of the space of possibilities or ‘state space’. (shrink)

The predictive processing literature contains at least two different versions of the framework with different theoretical resources at their disposal. One version appeals to so-called optimistic priors to explain agents’ motivation to act (call this optimistic predictive processing). A more recent version appeals to expected free energy minimization to explain how agents can decide between different action policies (call this preference predictive processing). The difference between the two versions has not been properly appreciated, and they are not sufficiently separated in (...) the literature. They constitute two different theories with strikingly different accounts of motivation and action. By reducing all desire-like constructs to belief-like constructs, optimistic predictive processing entails a substantial revision of standard accounts of motivation and action in philosophy and cognitive science. By contrast, preference predictive processing introduces desire-like constructs that play Humean motivational roles in the explanation of action. In this Humean iteration, predictive processing resembles other prominent computational frameworks implementing a distinction between beliefs and desires, such as reinforcement learning and Bayesian decision theory. Ultimately, predictive processing faces a dilemma between parsimony of mental constructs and completeness of its explanations of agency and the mind. (shrink)

This paper aims to contribute to the existing literature on normative cognition by providing a lineage explanation of human social norm psychology. This approach builds upon theories of goal-directed behavioral control in the reinforcement learning and control literature, arguing that this form of control defines an important class of intentional normative mental states that are instrumental in nature. I defend the view that great ape capacities for instrumental reasoning and our capacity (or family of capacities) for shared intentionality (...) coadapted to each other and argue that the evolution of this capacity has allowed the representation of social norms and the emergence of our capacity for normative guidance. (shrink)

We discuss the evolutionary transition from animals with limited experiencing to animals with unlimited experiencing and basic consciousness. This transition was, we suggest, intimately linked with the evolution of associative learning and with flexible reward systems based on, and modifiable by, learning. During associative learning, new pathways relating stimuli and effects are formed within a highly integrated and continuously active nervous system. We argue that the memory traces left by such new stimulus-effect relations form dynamic, flexible, and (...) varied global sensory states, which we call categorizing sensory states . These CSSs acquired a function: they came to act as internal “evaluators” and led to positive and negative reinforcement of new behavior. They are therefore the simplest, distinct, first-person motivational states that an animal can have. They constitute what we call basic consciousness, and are the hallmark of animals that can experience. Since associative learning has been found in many invertebrate taxa that first appeared during the Cambrian era, we propose that the processes underlying basic consciousness are phylogenetically ancient, and that their emergence may have fueled the Cambrian explosion. (shrink)

The Philippine educational system established alternative delivery modes of education, such as the Modified Work and Study Program, to eradicate student dropout incidence. This phenomenological study aims to probe the challenges and explore the coping mechanism of MWSP students in studying mathematics. The study was conducted at a public high school offering the program. Six students were chosen to participate in the study. Focus group discussion was utilized for the data gathering of the study. Results showed that mediocrity of resources (...) and conspicuous learning environment were the leading challenges that MWSP students encounter. Moreover, the coping mechanisms include good classroom management, utilization of concrete examples, time scheduling, steward support, motivational encouragement, and recreational and body relaxation. It is recommended that MWSP mathematics teachers integrate innovative strategies, provide more concrete contextualized examples, intensify the development of learning kits, and reinforce teaching and learning through supplementary materials. (shrink)

Discipline and interest are aspects of parenting that affect children’s behavior and academic performance. This study explores the parents’ disciplinary strategies to promote children’s learning interests. The researchers conducted this study due to society’s issue and observation that elementary pupils have a low interest in learning and manifested inappropriate behaviors inside and outside the classroom. The participants were the ten selected parents with children enrolled in 5th Grade in one of the private Catholic elementary schools in Pagadian City, (...) Philippines. A single case study was utilized. It was found that parents experienced challenges in disciplining their children to promote learning interests due to different kinds of behaviors that the children manifest. The participants identified the following behaviors: Too Much Screen Time in Playing Mobile Games and Being Disrespectful. Despite the challenges they have encountered, they were able to overcome those by implementing positive discipline, offering the problem to God, and effective communication to build good relationships with their children. It is important to be gentle and firm in implementing discipline to promote the learning interest of the pupils. Positive discipline inside the classroom allows the pupils to develop skills of responsibility, resiliency, and proper behavior with the use of positive and negative reinforcements that will motivate them to be interested in learning and excel academically. (shrink)

Context — Children and adults with psychopathic traits and conduct or oppositional defiant disorder demonstrate poor decision making and are impaired in reversal learning. However, the neural basis of this impairment has not previously been investigated. Furthermore, despite high comorbidity of psychopathic traits and attention deficit/hyperactivity disorder, to our knowledge, no research has attempted to distinguish neural correlates of childhood psychopathic traits and attention-deficit/hyperactivity disorder. Objective—To determine the neural regions that underlie the reversal learning impairments in children with (...) psychopathic traits plus conduct or oppositional defiant disorder. Design — Case-control study. Setting — Government clinical research institute. Participants — Forty-two adolescents aged 10 to 17 years: 14 with psychopathic traits and oppositional defiant disorder or conduct disorder, 14 with attention-deficit/hyperactivity disorder only, and 14 healthy controls. Main Outcome Measure — Blood oxygenation level–dependent signal as measured via functional magnetic resonance imaging during a probabilistic reversal task. Results — Children with psychopathic traits showed abnormal responses within the ventromedial prefrontal cortex (Brodmann area 10) during punished reversal errors compared with children wit hattention deficit/hyperactivity disorder and healthy children (P < .05 corrected for multiple comparisons). Conclusions — To our knowledge, this study provides the first evidence of abnormal ventromedial prefrontal cortex responsiveness in children with psychopathic traits and demonstrates this dysfunction was not attributable to comorbid attention-deficit/hyperactivity disorder. These findings suggest that reversal learning impairments in patients with developmental psychopathic traits relate to abnormal processing of reinforcement information. (shrink)

This essay examines the ontological and epistemological implications of artificial intelligence (AI) through the lens of posthumanist philosophy, integrating the works of Deleuze, Foucault, Haraway, and others with contemporary advancements in computational theory. It introduces concepts such as negative augmentation, praxes of revealing, and desedimentation as well as expands on concepts such as affirmative cartographies, the ethics of alterity, and planes of immanence to critique anthropocentric frameworks of identity, cognition, and agency. By situating AI within relational ontologies, the essay redefines (...) AI systems as dynamic assemblages whose subjectivities emerge through networks of interaction and co-creation, challenging traditional dichotomies of human versus machine and subject versus object. The analysis focuses on the spatial and geometric foundations of AI, contrasting Euclidean and non-Euclidean frameworks to explore how optimization processes, adversarial generative models, and reinforcement learning shape the epistemic assumptions of computational systems. It positions these systems within a praxis of revealing, emphasizing the generative potential of constraints and absences in fostering new modes of understanding. This essay advocates for a shift in AI ethics and safety discourse, proposing a posthumanist framework that prioritizes interconnectivity, plurality, and the emergent capacities of machine intelligence. By interrogating the phenomenology of AI systems and their co-constitutive relationships with human and non-human actors, it outlines a transformative vision for AI as an active participant in reconfiguring ontological possibilities and advancing epistemic pluralism in the digital age. (shrink)

Because culture plays a role in determining the aesthetic merit of a work of art, intrinsically similar works can have different aesthetic merit when assessed in different cultures. This paper argues that a form of aesthetic hedonism is best placed to account for this relativity of aesthetic value. This form of hedonism is based on a functional account of aesthetic pleasure, according to which it motivates and enables mental engagement with artworks, and an account of pleasure-learning, in which it (...) reinforces the appreciation of culture specific ways of engaging with art. (shrink)

Ever since 1988, Vietnam has successfully diversified and multilateralised its relationships, whilst placing a strong degree of focus on integration into the international political economy. This multidirectional foreign policy is designed to contribute to a peaceful international environment and a stable domestic one in order to promote economic growth and build up the aggregate strength of the country. At the same time, it is designed to boost the country’s autonomy, protect its sovereignty and territorial integrity, as well as hedge against (...) potential threats. This multidirectional foreign policy has contributed significantly to the economic reformist agenda, formally initiated at the 1986 Sixth Party Congress, otherwise known as Doi Moi. Additionally, it has boosted Vietnam’s international profile and reversed the diplomatic isolation it faced as the Cold War came to a close. As such, multidirectionalism has become a fundamental aspect of the Doi Moi process. This thesis traces the learning process since multidirectionalism’s inception and argues that this learning process, along with the economic benefits reaped, have evolved into positive memories for the Vietnamese Communist Party. This means multidirectionalism continues to be reinforced and as a result, these memories shape Vietnam’s continued expansion of that policy. Additionally, this thesis also offers a conceptual definition to the term multidirectionalism as well as explores the mechanisms through which Vietnam implements this policy. (shrink)

Some worry that advanced artificial agents may resist being shut down. The Incomplete Preferences Proposal (IPP) is an idea for ensuring that doesn't happen. A key part of the IPP is using a novel 'Discounted REward for Same-Length Trajectories (DREST)' reward function to train agents to (1) pursue goals effectively conditional on each trajectory-length (be 'USEFUL'), and (2) choose stochastically between different trajectory-lengths (be 'NEUTRAL' about trajectory-lengths). In this paper, we propose evaluation metrics for USEFULNESS and NEUTRALITY. We use a (...) DREST reward function to train simple agents to navigate gridworlds, and we find that these agents learn to be USEFUL and NEUTRAL. Our results thus suggest that DREST reward functions could also train advanced agents to be USEFUL and NEUTRAL, and thereby make these advanced agents useful and shutdownable. (shrink)

The first decade of this century has seen the nascency of the first mathematical theory of general artificial intelligence. This theory of Universal Artificial Intelligence (UAI) has made significant contributions to many theoretical, philosophical, and practical AI questions. In a series of papers culminating in book (Hutter, 2005), an exciting sound and complete mathematical model for a super intelligent agent (AIXI) has been developed and rigorously analyzed. While nowadays most AI researchers avoid discussing intelligence, the award-winning PhD thesis (Legg, 2008) (...) provided the philosophical embedding and investigated the UAI-based universal measure of rational intelligence, which is formal, objective and non-anthropocentric. Recently, effective approximations of AIXI have been derived and experimentally investigated in JAIR paper (Veness et al. 2011). This practical breakthrough has resulted in some impressive applications, finally muting earlier critique that UAI is only a theory. For the first time, without providing any domain knowledge, the same agent is able to self-adapt to a diverse range of interactive environments. For instance, AIXI is able to learn from scratch to play TicTacToe, Pacman, Kuhn Poker, and other games by trial and error, without even providing the rules of the games. These achievements give new hope that the grand goal of Artificial General Intelligence is not elusive. This article provides an informal overview of UAI in context. It attempts to gently introduce a very theoretical, formal, and mathematical subject, and discusses philosophical and technical ingredients, traits of intelligence, some social questions, and the past and future of UAI. (shrink)

Social norms are commonly understood as rules that dictate which behaviors are appropriate, permissible, or obligatory in different situations for members of a given community. Many researchers have sought to explain the ubiquity of social norms in human life in terms of the psychological mechanisms underlying their acquisition, conformity, and enforcement. Existing theories of the psychology of social norms appeal to a variety of constructs, from prediction-error minimization, to reinforcement learning, to shared intentionality, to domain-specific adaptations for norm (...) acquisition. In this paper, we propose a novel methodological and conceptual framework for the cognitive science of social norms that we call normative pluralism. We begin with an analysis of the explanatory aims of the cognitive science of social norms. From this analysis, we derive a recommendation for a reformed conception of its explanandum: a minimally psychological construct that we call normative regularities. Our central empirical proposal is that the psychological underpinnings of social norms are most likely realized by a heterogeneous set of cognitive, motivational, and ecological mechanisms that vary between norms and between individuals, rather than by a single type of process or distinctive norm system. This pluralistic approach, we suggest, offers a methodologically sound point of departure for a fruitful and rigorous science of social norms. (shrink)

I explain and motivate the shutdown problem: the problem of creating artificial agents that (1) shut down when a shutdown button is pressed, (2) don’t try to prevent or cause the pressing of the shutdown button, and (3) otherwise pursue goals competently. I then propose a solution: train agents to have incomplete preferences. Specifically, I propose that we train agents to lack a preference between every pair of different-length trajectories. I suggest a way to train such agents using reinforcement (...) learning: we give the agent lower reward for repeatedly choosing same-length trajectories. (shrink)

What is cognition? Equivalently, what is cognition good for? Or, what is it that would not be but for human cognition? But for human cognition, there would not be science. Based on this kinship between individual cognition and collective science, here we put forward Isbell conjugacy---the adjointness between objective geometry and subjective algebra---as a scientific method for developing cognitive science. We begin with the correspondence between categorical perception and category theory. Next, we show how the Gestalt maxim is subsumed by (...) the mathematical construct of colimit, a generalization of summation. The universal mapping property definitions of mathematical constructs, by virtue of being the best with respect to the universe of discourse, can be learned using reinforcement learning algorithms, which raises the possibility of abstracting the architecture of mathematics by artificial intelligence. Subsequently, we present naturality (to be contrasted with miracles), understood as 'Becoming consistent with Being', which governs the transformations of both things and their theories, as the zeroth law of change. Furthermore, the contrast---physical [mechanism] vs. biological [organism]---is smoothed via natural transformation, wherein transformations are respectful of the cohesion of the objects transformed. In closing, upon recognizing the scientific value of learning difficult-to-master differential calculus by physicists, of learning a strange four-letter language by biologists, and of learning the grammar of our respective mother tongues, we make a case for learning the theory of naturality / category theory for developing cognitive science. (shrink)

This paper investigates the nature of dispositional properties in the context of artificial intelligence systems. We start by examining the distinctive features of natural dispositions according to criteria introduced by McGeer (2018) for distinguishing between object-centered dispositions (i.e., properties like ‘fragility’) and agent-based abilities, including both ‘habits’ and ‘skills’ (a.k.a. ‘intelligent capacities’, Ryle 1949). We then explore to what extent the distinction applies to artificial dispositions in the context of two very different kinds of artificial systems, one based on rule-based (...) classical logic and the other on reinforcement learning. Here we defend three substantive claims. First, we argue that artificial systems are not equal in the kinds of dispositional properties they instantiate. In particular, we show that logical systems instantiate merely object-centered dispositions whereas reinforcement learning systems allow for the instantiation of agent-based abilities. Second, we explore the similarities and differences between the agent-centered abilities of artificial systems and those of humans, especially as relates to the important distinction made in the human case between habits and skills/intelligent capacities. The upshot is that the agent-centered abilities of truly intelligent artificial systems are distinctive enough to constitute a third type of agent-based ability — blended agent-based ability — raising substantial questions as to how we understand the nature of their agency. Third, we explore one aspect of this problem, focussing on whether systems of this type are properly considered ‘responsible agents’, at least in some contexts and for some purposes. The ramifications of our analysis will turn out to be directly relevant to various ethical concerns of artificial intelligence. (shrink)

Quilty-Dunn et al. argue that deep convolutional neural networks (DCNNs) optimized for image classification exemplify structural disanalogies to human vision. A different kind of artificial vision – found in reinforcement-learning agents navigating artificial three-dimensional environments – can be expected to be more human-like. Recent work suggests that language-like representations substantially improves these agents’ performance, lending some indirect support to the language-of-thought hypothesis (LoTH).

In this paper, we model a relational notion of subjectivity by means of two experiments in subjective computing. The goal is to determine to what extent a cognitive and social robot can be regarded to act subjectively. The system was implemented as a reinforcement learning agent with a coaching function. To analyze the robotic agent we used the method of levels of abstraction in order to analyze the agent at four levels of abstraction. At one level the agent (...) is described in mentalistic or subjective language respectively. By mapping this mentalistic to an algorithmic, functional, and relational level, we can show to what extent the agent behaves subjectively as we make use of a relational concept of subjectivity that draws upon the relations that hold between the agent and its environment. According to a relational notion of subjectivity, an agent is supposed to be subjective if it exhibits autonomous relations to itself and others, i.e. the agent is not fully determined by a given input but is able to operate on its input and decide what to do with it. This theoretical notion is confirmed by the technical implementation of self-referentiality and social interaction in that the agent shows improved behavior compared to agents without the ability of subjective computing. On the one hand, a relational concept of subjectivity is confirmed, whereas on the other hand, the technical framework of subjective computing is being theoretically founded. (shrink)

The potential for Large Language Models (LLMs) to attain technological singularity—the point at which artificial intelligence (AI) surpasses human intellect and autonomously improves itself—is a critical concern in AI research. This paper explores the feasibility of current LLMs achieving singularity by examining the philosophical and practical requirements for such a development. We begin with a historical overview of AI and intelligence amplification, tracing the evolution of LLMs from their origins to state-of-the-art models. We then proposes a theoretical framework to assess (...) whether existing LLM technologies could satisfy the conditions for singularity, with a focus on Recursive Self-Improvement (RSI) and autonomous code generation. We integrate key component technologies, such as Reinforcement Learning from Human Feedback (RLHF) and Direct Preference Optimization (DPO), into our analysis, illustrating how these could enable LLMs to independently enhance their reasoning and problem-solving capabilities. By mapping out a potential singularity model lifecycle and examining the dynamics of exponential growth models, we elucidate the conditions under which LLMs might self-replicate and rapidly escalate their intelligence. We conclude with a discussion of the ethical and safety implications of such developments, underscoring the need for responsible and controlled advancement in AI research to mitigate existential risks. Our work aims to contribute to the ongoing dialogue on the future of AI and the critical importance of proactive measures to ensure its beneficial development. (shrink)

The goal of creating Artificial General Intelligence (AGI) – or in other words of creating Turing machines (modern computers) that can behave in a way that mimics human intelligence – has occupied AI researchers ever since the idea of AI was first proposed. One common theme in these discussions is the thesis that the ability of a machine to conduct convincing dialogues with human beings can serve as at least a sufficient criterion of AGI. We argue that this very ability (...) should be accepted also as a necessary condition of AGI, and we provide a description of the nature of human dialogue in particular and of human language in general against this background. We then argue that it is for mathematical reasons impossible to program a machine in such a way that it could master human dialogue behaviour in its full generality. This is (1) because there are no traditional explicitly designed mathematical models that could be used as a starting point for creating such programs; and (2) because even the sorts of automated models generated by using machine learning, which have been used successfully in areas such as machine translation, cannot be extended to cope with human dialogue. If this is so, then we can conclude that a Turing machine also cannot possess AGI, because it fails to fulfil a necessary condition thereof. At the same time, however, we acknowledge the potential of Turing machines to master dialogue behaviour in highly restricted contexts, where what is called “narrow” AI can still be of considerable utility. (shrink)

The aim of this work is to move from the foreign dominated to the self-dominated by encouraging people to draw their own conclusions with the help of own rational consideration. Here a room as an environment that is encouraging innovation, which can be denoted as “Innovation Lab”, and making processes as can be regarded as “Smart Lab” is an essential base. The question related to this generalized self-organizational learning method investigated in our paper is how a UVC, which is (...) a room that connects people from different physical places to one synchronous and virtual perceivable place, which is built on these preconditions, can be operated both resource and learning-efficient for both the course participants and the educational organization. A practical approach of implementing a virtual classroom concept, including informative tutorial-feedback, is developed conceptually that also accounts for and implements the results of reinforcement machine-learning methods in AI applications. The difference that makes the difference is gained by reimplementing the AI tools in an AI instrument, in a “Smart Lab” environment and that in the teaching environment. By means of this, a cascaded feedback-loop system is informally installed, which gains feedback at different levels of abstraction. By this learning on each stage, in a collaborative and together decentralized and sequential fashion takes place, as the selforganizational implementations lead implicitly, also by means of the in the course implemented tools, to increasingly self-control. As such in the course, a tool is implemented, as generalizations by means of reinforcement learnings are to be emergently foreseen by this method, which goes beyond the tools, that have already been implemented before. This AI-enhanced learning coevolution shall then, predictively, as well increase the potential of the course participants as the educational organization according to the Wittgensteinean parable: A ladder leading into a selfly-organized future. (shrink)

We consider how an epistemic network might self-assemble from the ritualization of the individual decisions of simple heterogeneous agents. In such evolved social networks, inquirers may be significantly more successful than they could be investigating nature on their own. The evolved network may also dramatically lower the epistemic risk faced by even the most talented inquirers. We consider networks that self-assemble in the context of both perfect and imperfect communication and compare the behaviour of inquirers in each. This provides a (...) step in bringing together two new and developing research programs, the theory of self-assembling games and the theory of network epistemology. (shrink)

Interactions between an intelligent software agent and a human user are ubiquitous in everyday situations such as access to information, entertainment, and purchases. In such interactions, the ISA mediates the user’s access to the content, or controls some other aspect of the user experience, and is not designed to be neutral about outcomes of user choices. Like human users, ISAs are driven by goals, make autonomous decisions, and can learn from experience. Using ideas from bounded rationality, we frame these interactions (...) as instances of an ISA whose reward depends on actions performed by the user. Such agents benefit by steering the user’s behaviour towards outcomes that maximise the ISA’s utility, which may or may not be aligned with that of the user. Video games, news recommendation aggregation engines, and fitness trackers can all be instances of this general case. Our analysis facilitates distinguishing various subcases of interaction, as well as second-order effects that might include the possibility for adaptive interfaces to induce behavioural addiction, and/or change in user belief. We present these types of interaction within a conceptual framework, and review current examples of persuasive technologies and the issues that arise from their use. We argue that the nature of the feedback commonly used by learning agents to update their models and subsequent decisions could steer the behaviour of human users away from what benefits them, and in a direction that can undermine autonomy and cause further disparity between actions and goals as exemplified by addictive and compulsive behaviour. We discuss some of the ethical, social and legal implications of this technology and argue that it can sometimes exploit and reinforce weaknesses in human beings. (shrink)

More philosophical effort is spent articulating evolutionary rationales for the development of belief-like capacities than for precursors of desires or preferences. Nobody, though, seriously expects naturally evolved minds to be disinterested epistemologists. We agree that world-representing states won’t pay their way without supporting capacities that prioritise from an organism’s available repertoire of activities in light of stored (and occurrent) information. Some concede that desire-like states would be one way of solving this problem. Taking preferences as my starting point instead of (...) belief-like states, I defend two conclusions. First, psychologically real preference states, which approximately token expected utilities, have a quite general evolutionary rationale. They are a solution to the problem of efficiently allocating capacities with incompatible uses. This argument is a version of the Environmental Complexity Thesis. Second, preferences can plausibly function and naturally evolve without belief-like states, even though the converse claim is incredible. Preferences, that is, can mediate between discriminations of occurrent states (‘internal’ or ‘external’) and the processes selecting activity without mediation by stored indicative representations. By tokening expected utilities of actions conditional on discriminated state, they can increase the rate at which the ‘right thing’ is done at appropriate times, and they can do this without the support of belief-like, world-representing states. Preferences, even incomplete and noisy sets of them, are a fuel for success that will tend to be favoured when environments are complex in ways that matter to an organism, and when the organisms have complex behavioural repertoires with heterogenous returns and costs. (shrink)

The paper proposes a synthesis between human scientists and artificial representation learning models as a way of augmenting epistemic warrants of realist theories against various anti-realist attempts. Towards this end, the paper fleshes out unconceived alternatives not as a critique of scientific realism but rather a reinforcement, as it rejects the retrospective interpretations of scientific progress, which brought about the problem of alternatives in the first place. By utilising adversarial machine learning, the synthesis explores possibility spaces of (...) available evidence for unconceived alternatives providing modal knowledge of what is possible therein. As a result, the epistemic warrant of synthesised realist theories should emerge bolstered as the underdetermination by available evidence gets reduced. While shifting the realist commitment away from theoretical artefacts towards modalities of the possibility spaces, the synthesis comes out as a kind of perspectival modelling. (shrink)

This chapter explores some features of pragmatic pluralism as an ethical perspective on climate change. It is inspired in part by Andrew Light’s work on climate diplomacy as U.S. Assistant Secretary of Energy for International Affairs, and by Bryan Norton’s environmental pragmatism, while drawing more explicitly than Light or Norton from classical pragmatist sources such as John Dewey. The primary aim of the chapter is to characterize, differentiate, and advance a general pragmatist approach to climate ethics. The main line of (...) argument is that we are suffering culturally from a sort of “moral jetlag” due in part to “moral fundamentalist” habits, and that a critical focus on pragmatic pluralism—in moral theory generally and climate ethics particularly—would be salutary for our recovery if philosophers are to speak more effectively to “wicked problems” in a way that aids public deliberation and social learning. Moral fundamentalist habits, and the monistic one-way assumption that unintentionally—but not blamelessly—exercises and unduly reinforces them, are obstacles to fostering habits of moral and political inquiry better suited to dealing with predicaments rapidly transforming our warming planet. (shrink)

Human behavior is an underlying cause for many of the ecological crises faced in the 21st century, and there is no escaping from the fact that widespread behavior change is necessary for socio-ecological systems to take a sustainable turn. Whilst making people and communities behave sustainably is a fundamental objective for environmental policy, behavior change interventions and policies are often implemented from a very limited non-systemic perspective. Environmental policy-makers and psychologists alike often reduce cognition ‘to the brain,’ focusing only to (...) a minor extent on how everyday environments systemically afford pro-environmental behavior. Symptomatic of this are the widely prevalent attitude–action, value–action or knowledge–action gaps, understood in this paper as the gulfs lying between sustainable thinking and behavior due to lack of affordances. I suggest that by adopting a theory of affordances as a guiding heuristic, environmental policy-makers are better equipped to promote policies that translate sustainable thinking into sustainable behavior, often self-reinforcingly, and have better conceptual tools to nudge our socio–ecological system toward a sustainable turn. Affordance theory, which studies the relations between abilities to perceive and act and environmental features, is shown to provide a systemic framework for analyzing environmental policies and the ecology of human behavior. This facilitates the location and activation of leverage points for systemic policy interventions, which can help socio–ecological systems to learn to adapt to more sustainable habits. Affordance theory is presented to be applicable and pertinent to technically all nested levels of socio–ecological systems from the studies of sustainable objects and households to sustainable urban environments, making it an immensely versatile conceptual policy tool. Finally, affordance theory is also discussed from a participatory perspective. Increasing the fit between local thinking and external behavior possibilities entails a deep understanding of tacit and explicit attitudes, values, knowledge as well as physical and social environments, best gained via inclusive and polycentric policy approaches. (shrink)

This selective review explores biologically inspired learning as a model for intelligent robot control and sensing technology on the basis of specific examples. Hebbian synaptic learning is discussed as a functionally relevant model for machine learning and intelligence, as explained on the basis of examples from the highly plastic biological neural networks of invertebrates and vertebrates. Its potential for adaptive learning and control without supervision, the generation of functional complexity, and control architectures based on self-organization is (...) brought forward. Learning without prior knowledge based on excitatory and inhibitory neural mechanisms accounts for the process through which survival-relevant or task-relevant representations are either reinforced or suppressed. The basic mechanisms of unsupervised biological learning drive synaptic plasticity and adaptation for behavioral success in living brains with different levels of complexity. The insights collected here point toward the Hebbian model as a choice solution for “intelligent” robotics and sensor systems. Keywords: Hebbian learning; synaptic plasticity; neural networks; self-organization; brain; reinforcement; sensory processing; robot control . (shrink)

Georg Henrik von Wright was not only the first interpreter of Wittgenstein, who argued that Spengler’s work had reinforced and helped Wittgenstein to articulate his view of life, but also the first to consider seriously that Wittgenstein’s attitude to his times makes him unique among the great philosophers, that the philosophical problems which Wittgenstein was struggling, indeed his view of the nature of philosophy, were somehow connected with features of our culture or civilization. -/- In this paper I draw inspiration (...) and courage from Von Wright’s insistence that trying to understand Wittgenstein in relation to his times is a philosophic task in its own right in order to probe into a relatively obscure region in Wittgenstein’s thought: his relation to the music of his times. It is a topic, on which Von Wright, and most other prominent Wittgenstein scholars, have said very little, but it is also one, which Wittgenstein himself attested was so important to him that he felt without it he was sure to be misunderstood. -/- I offer textual and historical evidence in support of my claim that, parallel to Wittgenstein’s exposure to Spengler’s Decline of the West in 1930, he was also introduced to the music theory of Heinrich Schenker, which helped him to articulate, partly by way of critique, a complex and unique position concerning the modern music of his times, which exhibits his rejection of what Von Wright later dubbed ‘the myth of progress’. As Von Wright observed in other regions of Wittgenstein’s work, he believed also with regards to the arts and to music in particular, neither in a brilliant future nor in the good old days. -/- I argue that Wittgenstein actually made a distinction between three kinds of modern music: (a) bad modern music, which is clearly a case of confusing means for ends, the hallmark of the myth of progress, as Von Wright observed; (b) vacuous modern music, which embodies some sort of diffidence, a difficulty to see through the omnipresence of what Von Wright called (following Habermas) a ‘colonialization’ of reified measures of progress; (c) good modern music, a paradoxical notion for Wittgenstein, which betokens the unlikely yet possible striving to penetrate through what appears as dissolution of the resemblances which unite this culture’s ways of life by rendering this condition as expressible and intransitively understandable. In the context of this third category, I offer an interpretation of Wittgenstein’s complex remarks on the music of Gustav Mahler, which palpably show that the problem of good modern music and the problem of philosophizing in the time of civilization were one and the same in Wittgenstein’s mind. -/- I conclude that, with regards to Von Wright’s own critical view of the modern myth of progress, we can learn from Wittgenstein that progress in the realm of art is closely aligned with the ideal of the perfection of man, yet transcending a social or political context. It is the ideal of cultural cohesion: affinity that the arts show to other human practices and cultural artifacts of its period. Wittgenstein’s tentative notion of good modern music (and its circumscription by his notion of the music of the future) may show its true colors when viewed in the context of Von Wright’s plea not to abandon work for progress as a critical task. (shrink)

This chapter discusses a form of pedagogy of reflection suggested to be defined as the dialogical-reflective professional-development school (DRPDS)
a framework that develops and empowers students by engaging them in a process of continual improvement, responding to diverse situations, providing stimuli for learning, and giving anchors for mediation. The pedagogy of reflection relates to dialogue not only from a theoretical historical context but also by way of example
that is, it offers empowering dialogues within the traditional teacher-training (...) framework. This chapter outlines the importance of the pedagogy of reflection in the multicultural educational space of the preservice education field in Israel, analyzing the first university PDS model. The pedagogy of reflection in the context of the educational dialogue of educators is outlined as a tool for student empowerment, achieved through a community of learners who dedicate space to the development of their whole personality within the profession, taking a moral stance toward the educational discourse, minimizing judgmentalism and prejudice, creating national/gender equality with the goal of examining the fundamental question of educational performance, and reinforcing their sense of organizational belonging within the system. In these contexts, the chapter is based on the elements of dialogical philosophy exemplified in the thought of Burbules, Nelson, Isaacs, Bohm, and Heckmann and the reflective basis of educational and organizational performance exemplified in the writings of van Manen. The chapter also presents two examples from a project in which teaching units based on dialogue and reflection were developed within a dialogic community that represents in its very being collective empowerment, the possibility of coping with problems that are too large for an individual to solve on his/her own, and an alternative to sealed and alienated organizations. (shrink)

This study aims to find out the factors that hinder students in learning Integrated Art Education. A descriptive research design was utilized in the conduct of the study. The researcher prepared a questionnaire with 15 closed-ended- questions that could be answered objectively. The study discovered that the students would learn more when they feel that they belong to a certain group. Interests in a subject also matter, the more you are interested in a particular subject, the more you will (...) learn the material. Parent support marks an impact on the student's interest in learning Arts. Therefore, personal interest in a subject must be given attention and managed properly. Parent's support also plays a vital role in reinforcing the students on what to pursue in order to maintain their motivation. (shrink)