As a way to address both ominous and ordinary threats of artificial intelligence, researchers have started proposing ways to stop an AI system before it has a chance to escape outside control and cause harm. A so-called “big red button” would enable human operators to interrupt or divert a system while preventing the system from learning that such an intervention is a threat. Though an emergency button for AI seems to make intuitive sense, that approach ultimately concentrates on the point (...) when a system has already “gone rogue” and seeks to obstruct interference. A better approach would be to make ongoing self-evaluation and testing an integral part of a system’s operation, diagnose how the system is in error and to prevent chaos and risk before they start. In this paper, we describe the demands that recent big red button proposals have not addressed, and we offer a preliminary model of an approach that could better meet them. We argue for an ethical core that consists of a scenario-generation mechanism and a simulation environment that are used to test a system’s decisions in simulated worlds, rather than the real world. This EC would be kept opaque to the system itself: through careful design of memory and the character of the scenario, the system’s algorithms would be prevented from learning about its operation and its function, and ultimately its presence. By monitoring and checking for deviant behavior, we conclude, a continual testing approach will be far more effective, responsive, and vigilant toward a system’s learning and action in the world than an emergency button which one might not get to push in time. (shrink)

After briefly discussing the relevance of the notions computation and implementation for cognitive science, I summarize some of the problems that have been found in their most common interpretations. In particular, I argue that standard notions of computation together with a state-to-state correspondence view of implementation cannot overcome difficulties posed by Putnam's Realization Theorem and that, therefore, a different approach to implementation is required. The notion realization of a function, developed out of physical theories, is then introduced as a replacement (...) for the notional pair computation-implementation. After gradual refinement, taking practical constraints into account, this notion gives rise to the notion digital system which singles out physical systems that could be actually used, and possibly even built. (shrink)

This paper argues against the moral Turing test as a framework for evaluating the moral performance of autonomous systems. Though the term has been carefully introduced, considered, and cautioned about in previous discussions :251–261, 2000; Allen and Wallach 2009), it has lingered on as a touchstone for developing computational approaches to moral reasoning :98–109, 2015). While these efforts have not led to the detailed development of an MTT, they nonetheless retain the idea to discuss what kinds of action and reasoning (...) should be demanded of autonomous systems. We explore the flawed basis of an MTT in imitation, even one based on scenarios of morally accountable actions. MTT-based evaluations are vulnerable to deception, inadequate reasoning, and inferior moral performance vis a vis a system’s capabilities. We propose verification—which demands the design of transparent, accountable processes of reasoning that reliably prefigure the performance of autonomous systems—serves as a superior framework for both designer and system alike. As autonomous social robots in particular take on an increasing range of critical roles within society, we conclude that verification offers an essential, albeit challenging, moral measure of their design and performance. (shrink)

The main claim of this paper is that notions of implementation based on an isomorphic correspondence between physical and computational states are not tenable. Rather, ``implementation'' has to be based on the notion of ``bisimulation'' in order to be able to block unwanted implementation results and incorporate intuitions from computational practice. A formal definition of implementation is suggested, which satisfies theoretical and practical requirements and may also be used to make the functionalist notion of ``physical realization'' precise. The upshot of (...) this new definition of implementation is that implementation cannot distinguish isomorphic bisimilar from non-isomporphic bisimilar systems anymore, thus driving a wedge between the notions of causal and computational complexity. While computationalism does not seem to be affected by this result, the consequences for functionalism are not clear and need further investigations. (shrink)

The main claim of this paper is that notions of implementation based on an isomorphic correspondence between physical and computational states are not tenable. Rather, ``implementation'' has to be based on the notion of ``bisimulation'' in order to be able to block unwanted implementation results and incorporate intuitions from computational practice. A formal definition of implementation is suggested, which satisfies theoretical and practical requirements and may also be used to make the functionalist notion of ``physical realization'' precise. The upshot of (...) this new definition of implementation is that implementation cannot distinguish isomorphic bisimilar from non-isomporphic bisimilar systems anymore, thus driving a wedge between the notions of causal and computational complexity. While computationalism does not seem to be affected by this result, the consequences for functionalism are not clear and need further investigations. (shrink)

A new computationalist view of the mind that takes into account real-world issues of embodiment, interaction, physical implementation, and semantics.

Even though morally competent artificial agents have yet to emerge in society, we need insights from empirical science into how people will respond to such agents and how these responses should inform agent design. Three survey studies presented participants with an artificial intelligence agent, an autonomous drone, or a human drone pilot facing a moral dilemma in a military context: to either launch a missile strike on a terrorist compound but risk the life of a child, or to cancel the (...) strike to protect the child but risk a terrorist attack. Seventy-two percent of respondents were comfortable making moral judgments about the AI in this scenario and fifty-one percent were comfortable making moral judgments about the autonomous drone. These participants applied the same norms to the two artificial agents and the human drone pilot. However, people ascribed different patterns of blame to humans and machines as a function of the agent’s decision of how to solve the dilemma. These differences in blame seem to stem from different assumptions about the agents’ embeddedness in social structures and the moral justifications those structures afford. Specifically, people less readily see artificial agents as embedded in social structures and, as a result, they explained and justified their actions differently. As artificial agents will perform many actions with moral significance, we must heed such differences in justifications and blame and probe how they affect our interactions with those agents. (shrink)

Making interactions between humans and artificial agents successful is a major goal of interaction design. The aim of this paper is to provide researchers conducting interaction studies a new framework for the evaluation of robot believability. By critically examining the ordinary sense of believability, we first argue that currently available notions of it are underspecified for rigorous application in an experimental setting. We then define four concepts that capture different senses of believability, each of which connects directly to an empirical (...) methodology. Finally, we show how this framework has been and can be used in the construction of interaction studies by applying it to our own work in human–robot interaction. (shrink)

Cooper et al. (this issue) develop an interactive activation model of spatial and imitative compatibilities that simulates the key results from Catmur and Heyes (2011) and thus conclude that both compatibilities are mediated by the same processes since their single model can predict all the results. Although the model is impressive, the conclusions are premature because they are based on an incomplete review of the relevant literature and because the model includes some questionable assumptions. Moreover, a competing model (Scheutz & (...) Bertenthal, 2012) is introduced that suggests the two compatibilities are not mediated by the same processes. We propose that more research is necessary before concluding that spatial and imitative compatibilities are mediated by the same processes. (shrink)

Although Ross & Spurrett (R&S) successfully fend off the threat of Kim's “supervenience argument” by showing that it conflates different notions of causation, their proposal for a dynamic systems answer to the mind-body problem is itself yet another supervenience claim in need of an explanation that justifies it. The same goes for their notion of “multiple supervenience.”.

This paper addresses ethical challenges posed by a robot acting as both a general type of system and a discrete, particular machine. Using the philosophical distinction between “type” and “token,” we locate type-token ambiguity within a larger field of indefinite robotic identity, which can include networked systems or multiple bodies under a single control system. The paper explores three specific areas where the type-token tension might affect human–robot interaction, including how a robot demonstrates the highly personalized recounting of information, how (...) a robot makes moral appeals and justifies its decisions, and how the possible need for replacement of a particular robot shapes its ongoing role. We also consider how a robot might regard itself as a replaceable token of a general robotic type and take extraordinary actions on that basis. For human–robot interaction robotic type-token identity is not an ontological problem that has a single solution, but a range of possible interactions that responsible design must take into account, given how people stand to gain and lose from the shifting identities social robots will present. (shrink)

Soft robots promise an exciting design trajectory in the field of robotics and human–robot interaction (HRI), promising more adaptive, resilient movement within environments as well as a safer, more sensitive interface for the objects or agents the robot encounters. In particular, tactile HRI is a critical dimension for designers to consider, especially given the onrush of assistive and companion robots into our society. In this article, we propose to surface an important set of ethical challenges for the field of soft (...) robotics to meet. Tactile HRI strongly suggests that soft-bodied robots balance tactile engagement against emotional manipulation, model intimacy on the bonding with a tool not with a person, and deflect users from personally and socially destructive behavior the soft bodies and surfaces could normally entice. (shrink)

Human instructors often refer to objects and actions involved in a task description using both linguistic and non-linguistic means of communication. Hence, for robots to engage in natural human-robot interactions, we need to better understand the various relevant aspects of human multi-modal task descriptions. We analyse reference resolution to objects in a data collection comprising two object manipulation tasks and find that 78.76% of all referring expressions to the objects relevant in Task 1 are verbally underspecified and 88.64% of all (...) referring expressions are verbally underspecified in Task 2. The data strongly suggests that a language processing module for robots must be genuinely multi-modal, allowing for seamless integration of information transmitted in the verbal and the visual channel, whereby tracking the speaker’s eye gaze and gestures as well as object recognition are necessary preconditions. (shrink)

Any discussion comparing different models with respect to their quality qua models must presuppose a notion of model, that is, what it is to be a model. While Webb provides seven criteria to assess the quality of various proposed biorobotic models, she does not clarify the very notion of “model of animal behavior” itself.