Interest in computerised central economic planning (CCEP) has seen a resurgence, as there is strong demand for an alternative vision to modern free (or not so free) market liberal capitalism. Given the close links of CCEP with what we would now broadly call artificial intelligence (AI)—e.g. optimisation, game theory, function approximation, machine learning, automated reasoning—it is reasonable to draw direct analogues and perform an analysis that would help identify what commodities and institutions we should see for a CCEP programme to (...) become successful. Following this analysis, we conclude that a CCEP economy would need to have a very different outlook from current market practices, with a focus on producing basic “interlinking” commodities (e.g. tools, processed materials, instruction videos) that consumers can use as a form of collective R &D. On an institutional level, CCEP should strive for the release of basic commodities that empower consumers by having as many alternative uses as possible, but also making sure that a baseline of basic necessities is widely available. (shrink)

Theories of rational decision making often abstract away from computational and other resource limitations faced by real agents. An alternative approach known as resource rationality puts such matters front and center, grounding choice and decision in the rational use of finite resources. Anticipated by earlier work in economics and in computer science, this approach has recently seen rapid development and application in the cognitive sciences. Here, the theory of rationality plays a dual role, both as a framework for normative assessment (...) and as a source of scientific hypotheses about how mental processes in fact work. The latter project, often called rational analysis, depends for its success on a fine-grained characterization of the computational problem facing a decision maker, which may in turn depend on realistic assumptions about what the relevant agent is like. As a consequence, resource rationality involves a delicate, but often fruitful interplay between the normative and the descriptive. (shrink)

Machine learning (ML) techniques have become pervasive across a range of different applications, and are now widely used in areas as disparate as recidivism prediction, consumer credit-risk analysis, and insurance pricing. Likewise, in the physical world, ML models are critical components in autonomous agents such as robotic surgeons and self-driving cars. Among the many ethical dimensions that arise in the use of ML technology in such applications, analyzing morally permissible actions is both immediate and profound. For example, there is the (...) potential for learned algorithms to become biased against certain groups. More generally, in so much that the decisions of ML models impact society, both virtually (e.g., denying a loan) and physically (e.g., driving into a pedestrian), notions of accountability, blame and responsibility need to be carefully considered. In this article, we advocate for a two-pronged approach ethical decision-making enabled using rich models of autonomous agency: on the one hand, we need to draw on philosophical notions of such as beliefs, causes, effects and intentions, and look to formalise them, as attempted by the knowledge representation community, but on the other, from a computational perspective, such theories need to also address the problems of tractable reasoning and (probabilistic) knowledge acquisition. As a concrete instance of this tradeoff, we report on a few preliminary results that apply (propositional) tractable probabilistic models to problems in fair ML and automated reasoning of moral principles. Such models are compilation targets for certain types of knowledge representation languages, and can effectively reason in service some computational tasks. They can also be learned from data. Concretely, current evidence suggests that they are attractive structures for jointly addressing three fundamental challenges: reasoning about possible worlds + tractable computation + knowledge acquisition. Thus, these seems like a good starting point for modelling reasoning robots as part of the larger ecosystem where accountability and responsibility is understood more broadly. (shrink)

Although virtue ethics has repeatedly been proposed as a suitable framework for the development of artificial moral agents, it has been proven difficult to approach from a computational perspective. In this work, we present the first technical implementation of artificial virtuous agents in moral simulations. First, we review previous conceptual and technical work in artificial virtue ethics and describe a functionalistic path to AVAs based on dispositional virtues, bottom-up learning, and top-down eudaimonic reward. We then provide the details of a (...) technical implementation in a moral simulation based on a tragedy of the commons scenario. The experimental results show how the AVAs learn to tackle cooperation problems while exhibiting core features of their theoretical counterpart, including moral character, dispositional virtues, learning from experience, and the pursuit of eudaimonia. Ultimately, we argue that virtue ethics provides a compelling path toward morally excellent machines and that our work provides an important starting point for such endeavors. (shrink)

Can humans get arbitrarily capable reinforcement learning agents to do their bidding? Or will sufficiently capable RL agents always find ways to bypass their intended objectives by shortcutting their reward signal? This question impacts how far RL can be scaled, and whether alternative paradigms must be developed in order to build safe artificial general intelligence. In this paper, we study when an RL agent has an instrumental goal to tamper with its reward process, and describe design principles that prevent instrumental (...) goals for two different types of reward tampering. Combined, the design principles can prevent reward tampering from being an instrumental goal. The analysis benefits from causal influence diagrams to provide intuitive yet precise formalizations. (shrink)

Human and automated tutors attempt to choose pedagogical activities that will maximize student learning, informed by their estimates of the student's current knowledge. There has been substantial research on tracking and modeling student learning, but significantly less attention on how to plan teaching actions and how the assumed student model impacts the resulting plans. We frame the problem of optimally selecting teaching actions using a decision-theoretic approach and show how to formulate teaching as a partially observable Markov decision process planning (...) problem. This framework makes it possible to explore how different assumptions about student learning and behavior should affect the selection of teaching actions. We consider how to apply this framework to concept learning problems, and we present approximate methods for finding optimal teaching actions, given the large state and action spaces that arise in teaching. Through simulations and behavioral experiments, we explore the consequences of choosing teacher actions under different assumed student models. In two concept-learning tasks, we show that this technique can accelerate learning relative to baseline performance. (shrink)

We consider a reinforcement learning setting where the learner is given a set of possible models containing the true model. While there are algorithms that are able to successfully learn optimal behavior in this setting, they do so without trying to identify the underlying true model. Indeed, we show that there are cases in which the attempt to find the true model is doomed to failure.

Why is interaction so simple? This article presents a theory of interaction based on the use of shared representations as “coordination tools” (e.g., roundabouts that facilitate coordination of drivers). By aligning their representations (intentionally or unintentionally), interacting agents help one another to solve interaction problems in that they remain predictable, and offer cues for action selection and goal monitoring. We illustrate how this strategy works in a joint task (building together a tower of bricks) and discuss its requirements from a (...) computational viewpoint. (shrink)

Curiosity and creativity are expressions of the trade-off between leveraging that with which we are familiar or seeking out novelty. Through the computational lens of reinforcement learning, we describe how formulating the value of information seeking and generation via their complementary effects on planning horizons formally captures a range of solutions to striking this balance.

Knowledge-based programs (KBPs) are a powerful notion for expressing action policies in which branching conditions refer to implicit knowledge and call for a deliberation task at execution time. However, branching conditions in KBPs cannot refer to possibly erroneous beliefs or to graded belief, such as “if my belief that φ holds is high then do some action α else perform some sensing action β”.

How can an incomplete and uncertain model of the environment be used to perceive, infer, decide and act efficiently? This is the challenge that both living and artificial cognitive systems have to face. Symbolic logic is, by its nature, unable to deal with this question. The subjectivist approach to probability is an extension to logic that is designed specifically to face this challenge. In this paper, we review a number of frequently encountered cognitive issues and cast them into a common (...) Bayesian formalism. The concepts we review are ambiguities, fusion, multimodality, conflicts, modularity, hierarchies and loops. First, each of these concepts is introduced briefly using some examples from the neuroscience, psychophysics or robotics literature. Then, the concept is formalized using a template Bayesian model. The assumptions and common features of these models, as well as their major differences, are outlined and discussed. (shrink)