We might worry that our seamless reliance on AI systems makes us prone to adopting the strange errors that these systems commit. One proposed solution is to design AI systems so that they are not phenomenally transparent to their users. This stops cognitive extension and the automatic uptake of errors. Although we acknowledge that some aspects of AI extension are concerning, we can address these concerns without discouraging transparent employment altogether. First, we believe that the potential danger should be put (...) into perspective – many unreliable technologies are unlikely to be used transparently precisely because they are unreliable. Second, an agent who transparently employs a resource may also reflect (opaquely) on its reliability. Finally, agents can rely on a process transparently and be yanked out of their transparent use when it turns unreliable. When an agent is responsive to the reliability of their process in this way, they have epistemically integrated it, and the beliefs they form with it are formed responsibly. This prevents the agent from automatically incorporating problematic beliefs. Responsible (and transparent) use of AI resources – and consequently responsible AI extension – is hence possible. We end the paper with several design and policy recommendations that encourage epistemic integration of AI-involving belief-forming processes. (shrink)

Phenomenal transparency was once widely believed to be necessary for cognitive extension. Recently, this claim has come under attack, with a new consensus coalescing around the idea that transparency is neither necessary for internal nor extended cognitive processes. We take these recent critiques as an opportunity to refine the concept of transparency relevant for cognitive extension. In particular, we highlight that transparency concerns an agent’s employment of a resource – and that such employment is compatible with an agent consciously apprehending (...) (or attending to) a resource. This means it is possible for an object to be transparent and opaque to an agent, even at a single moment in time. Once we understand transparency in this way, the detractors’ claims lose their bite, and existing arguments for transparency’s necessity for cognitive extension return to apply with full force. (shrink)

Virtue reliabilists argue that an agent can only gain knowledge if she responsibly employs a reliable belief-forming process. This in turn demands that she is either aware that her process is reliable or is sensitive to her process’s reliability in some other way. According to a recent argument in the philosophy of mind, sometimes a cognitive mechanism (i.e. precision estimation) can ensure that a belief-forming process is only employed when it’s reliable. If this is correct, epistemic responsibility can sometimes be (...) explained entirely on the subpersonal level. In this paper, I argue that the mechanism of precision estimation–the alleged new variety of epistemic responsibility–is a more ubiquitous phenomenon than epistemic responsibility. I show that precision estimation operates at levels that are not always concerned with the epistemic domain. Lastly, I broaden this argument to explain how all subpersonal epistemologies are likely to fall prey to the problem of demarcating cognitive agency and the problem of attributing beliefs. (shrink)

Virtue reliabilism provides an account of epistemic integration that explains how a reliable-belief forming process can become a knowledge-conducive ability of one’s cognitive character. The univocal view suggests that this epistemic integration can also explain how an external process can extend one’s cognition into the environment. Andy Clark finds a problem with the univocal view. He claims that cognitive extension is a wholly subpersonal affair, whereas the epistemic integration that virtue reliabilism puts forward requires personal-level agential involvement. To adjust the (...) univocal view, Clark recommends a subpersonal epistemic integration account that also paves way for a wholly subpersonal epistemology. Accordingly, an epistemic agent can take responsibility for her reliable belief-forming process by way of entirely subpersonal mechanisms. The aim of this thesis is to argue against a subpersonal epistemology and the need for it. First, I bring into question the conditions that motivate extended cognition: the so-called ‘glue and trust’ requirements and the functional parity principle. Neither of these conditions demands that extension should be understood in entirely subpersonal terms. On the contrary, the glue and trust conditions suggest that agents should personally and actively engage with their external vehicles to extend cognition. Further, I consider an important disparity between the two kinds of integration. Integration that prompts extension can happen immediately, but the integration that makes a reliable belief-forming process knowledge-conducive is almost always a slow process. In light of this and other similar inconsistencies between the two integration accounts, I conclude that a type of univocal view is difficult to defend. And, since that type of univocal view is the main reason for a subpersonal epistemology, the need for it does not arise. Next, I locate Clark’s main motivation for his subpersonal epistemology in his account of the predictive brain. While the predictive brain provides useful insight into how an external process might extend one’s cognition, I show it runs into problems when it tries to account for epistemic integration. Then, I explore how the concept of epistemic defeaters informs epistemic integration. Agents have to meet a specific no-defeater condition to be sensitive to the reliability of their belief-forming processes and to employ them responsibly. In an entirely subpersonal epistemology, defeasibility theory has no means to explain how an agent is rendered sensitive to the reliability of her process. Subpersonal mechanisms can become sensitive to the new process, but that does not allow the agent to employ said process responsibly. Finally, I use an AI case, based on Amazon Alexa, to argue that all accounts of epistemic integration ought to explicitly describe how one’s subpersonal mechanisms link to one’s whole person, even if the link is weak and indirect. Without this connection between one’s subpersonal mechanisms and one’s person, it can become difficult to ascribe beliefs to the agent. This becomes apparent when Alexa leads to ever-increasing layers of cognitive, mental, and agential extension. (shrink)