Nudge is a popular public policy tool that harnesses well-known biases in human judgement to subtly guide people’s decisions, often to improve their choices or to achieve some socially desirable outcome. Thanks to recent developments in artificial intelligence (AI) methods new possibilities emerge of how and when our decisions can be nudged. On the one hand, algorithmically personalized nudges have the potential to vastly improve human daily lives. On the other hand, blindly outsourcing the development and implementation of nudges to (...) “black box” AI systems means that the ultimate reasons for why such nudges work, that is, the underlying human cognitive processes that they harness, will often be unknown. In this paper, we unpack this concern by considering a series of examples and case studies that demonstrate how AI systems can learn to harness biases in human judgment to reach a specified goal. Drawing on an analogy in a philosophical debate concerning the methodology of economics, we call for the need of an interdisciplinary oversight of AI systems that are tasked and deployed to nudge human behaviours. (shrink)

We cooperate with other people despite the risk of being exploited or hurt. If future artificial intelligence (AI) systems are benevolent and cooperative toward us, what will we do in return? Here we show that our cooperative dispositions are weaker when we interact with AI. In nine experiments, humans interacted with either another human or an AI agent in four classic social dilemma economic games and a newly designed game of Reciprocity that we introduce here. Contrary to the hypothesis that (...) people mistrust algorithms, participants trusted their AI partners to be as cooperative as humans. However, they did not return AI's benevolence as much and exploited the AI more than humans. These findings warn that future self-driving cars or co-working robots, whose success depends on humans' returning their cooperativeness, run the risk of being exploited. This vulnerability calls not just for smarter machines but also better human-centered policies. (shrink)

From playing basketball to ordering at a food counter, we frequently and effortlessly coordinate our attention with others towards a common focus: we look at the ball, or point at a piece of cake. This non-verbal coordination of attention plays a fundamental role in our social lives: it ensures that we refer to the same object, develop a shared language, understand each other’s mental states, and coordinate our actions. Models of joint attention generally attribute this accomplishment to gaze coordination. But (...) are visual attentional mechanisms sufficient to achieve joint attention, in all cases? Besides cases where visual information is missing, we show how combining it with other senses can be helpful, and even necessary to certain uses of joint attention. We explain the two ways in which non-visual cues contribute to joint attention: either as enhancers, when they complement gaze and pointing gestures in order to coordinate joint attention on visible objects, or as modality pointers, when joint attention needs to be shifted away from the whole object to one of its properties, say weight or texture. This multisensory approach to joint attention has important implications for social robotics, clinical diagnostics, pedagogy and theoretical debates on the construction of a shared world. (shrink)

What if a blind person could 'see' with her ears? Thanks to Sensory Substitution Devices (SSDs), blind people now have access to out-of-reach objects, a privilege reserved so far for the sighted. In this paper, we show that the philosophical debates have fundamentally been mislead to think that SSDs should be fitted among the existing senses or that they constitute a new sense. Contrary to the existing assumption that they get integrated at the sensory level, we present a new thesis (...) according to which they are not sensory, and get vertically integrated on the top of existing sensory abilities, from which they should be theoretically distinguished. (shrink)

Many policies and ethical guidelines recommend developing “trustworthy AI”. We argue that developing morally trustworthy AI is not only unethical, as it promotes trust in an entity that cannot be trustworthy, but it is also unnecessary for optimal calibration. Instead, we show that reliability, exclusive of moral trust, entails the appropriate normative constraints that enable optimal calibration and mitigate the vulnerability that arises in high-stakes hybrid decision-making environments, without also demanding, as moral trust would, the anthropomorphization of AI and thus (...) epistemically dubious behavior. The normative demands of reliability for inter-agential action are argued to be met by an analogue to procedural metacognitive competence (i.e., the ability to evaluate the quality of one’s own informational states to regulate subsequent action). Drawing on recent empirical findings that suggest providing reliability scores (e.g., F1-scores) to human decision-makers improves calibration in the AI system, we argue that reliability scores provide a good index of competence and enable humans to determine how much they wish to rely on the system. (shrink)

Cognitive states, such as beliefs, desires and intentions, may influence how we perceive people and objects. If this is the case, are those influences worse when they occur implicitly rather than explicitly? Here we show that cognitive penetration in perception generally involves an implicit component. First, the process of influence is implicit, making us unaware that our perception is misrepresenting the world. This lack of awareness is the source of the epistemic threat raised by cognitive penetration. Second, the influencing state (...) can be implicit, though it can also be or become explicit. Being unaware of the content of the influencing state, we argue, does not make as much difference to the epistemic threat as it does to the epistemic responsibility of the agent. Implicit influencers cannot be examined for their accuracy and justification, and cannot be voluntarily accepted by the perceiver. Conscious awareness, however, is not sufficient for attributing blame to the agent. An equally important condition is the degree of control that they can exercise to change the contents that influence perception or stop their influence. Here we suggest that such control can also result from social influence, and that cognitive penetrability of perception is therefore also a social issue. (shrink)

A common sense view is illustrated by Doubting Thomas, and surfaces in many philosophical and psychological writings : Touching is better than seeing. But can we make sense of this privilege? We rule out that it could mean that touch is more informative than vision, more ‘objective’ or more directly in contact with reality. Instead, we propose that touch offers not a perceptual, but a metacognitive advantage: touch is not more objective than vision but rather provides comparatively higher subjective certainty.

Humans coordinate their focus of attention with others, either by gaze following or prior agreement. Though the effects of joint attention on perceptual and cognitive processing tend to be examined in purely visual environments, they should also show in multisensory settings. According to a prevalent hypothesis, joint attention enhances visual information encoding and processing, over and above individual attention. If two individuals jointly attend to the visual components of an audiovisual event, this should affect the weighing of visual information during (...) multisensory integration. We tested this prediction in this preregistered study, using the well-documented sound-induced flash illusions, where the integration of an incongruent number of visual flashes and auditory beeps results in a single flash being seen as two (fission illusion) and two flashes as one (fusion illusion). Participants were asked to count flashes either alone or together, and expected to be less prone to both fission and fusion illusions when they jointly attended to the visual targets. However, illusions were as frequent when people attended to the flashes alone or with someone else, even though they responded faster during joint attention. Our results reveal the limitations of the theory that joint attention enhances visual processing as it does not affect temporal audiovisual integration. (shrink)