Current Events

Late in his career, Daniel Dennett changed his mind about minds in a way that deserves careful attention. In "Brains Made of Brains" (2017), Dennett expresses regret about aspects of his earlier "homuncular functionalism" program, which explained cognitive tasks by decomposing them into simpler sub-tasks performed by less sophisticated homunculi. While maintaining his functionalism, Dennett now suggests that genuine minds require a different kind of architecture: one built from parts that themselves have simpler forms of caring and agency. This paper examines and develops this "desire homuncularism" view, which holds both that genuine minds require genuine caring, and that such caring requires parts with simpler cares. I argue that this view helps explain persistent intuitions about which systems have minds, and has significant implications for both AI sentience and existential risk from AI. Drawing on work by Terrence Deacon, I develop two key arguments for desire homuncularism: First, systems with distributed caring exhibit deeper integration, as evidenced by how changes to their parts affect their overall goals. Second, such systems achieve a richer form of conative reference - their goals are more firmly grounded in the world and harder to "spoof." These considerations suggest modern AI systems, despite their impressive capabilities, may be what Keith Frankish calls "cognitively rich but conatively bankrupt." While they can process information in remarkably sophisticated ways, they lack the kind of distributed, bottom-up caring that characterizes genuine minds. This analysis points to a possible path forward in AI development: we may be able to create highly capable AI systems that neither warrant significant moral concern (due to lack of genuine caring) nor pose existential risks (due to lack of genuine agency). While speculative, this view suggests that the old science fiction dream of helpful but largely "mindless" robots may be achievable after all.

Cognitive Work Analysis (CWA) is a conceptual framework that allows for analysis of all factors that affect human-system interactions. The products of this system of analyses can be directly transformed into design requirements for information systems. CWA is an inherently holistic approach that simultaneously examines the environmental, organizational, and social activities, as well as individual dimensions. CWA consists of six stages, which focus on the reasons that a worker may prefer one cognitive strategy over another or may transition between strategies during execution of a cognitive process, and identifies the skills-, rules, or knowledge (SRK) modes of cognition. The designer may choose to encourage and induce one cognitive mode over another as dictated by the situation (e.g., in an emergency response, skill-based performance is preferred). This presentation reviews the genesis and theoretical underpinnings of CWA and the specific challenges in its use and implementation, pointing to still open research questions.

TBA

TBA

TBA

TBA