Thinking Fast And Slow In Human-Centered AI
| Autor: | Dahlgren Lindström, Adam, Mackay, Wendy E., Dignum, Virginia |
|---|---|
| Přispěvatelé: | Umeå University, Extreme Situated Interaction (EX-SITU), Inria Saclay - Ile de France, Institut National de Recherche en Informatique et en Automatique (Inria)-Institut National de Recherche en Informatique et en Automatique (Inria)-Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Interaction avec l'Humain (IaH), Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), AAAI Association for the Advancement of Artificial Intelligence, AAAI |
| Jazyk: | angličtina |
| Rok vydání: | 2022 |
| Předmět: | |
| Zdroj: | FSS 2022-AAAI Fall symposium Series-Thinking Fast and Slow and Other Cognitive Theories in AI FSS 2022-AAAI Fall symposium Series-Thinking Fast and Slow and Other Cognitive Theories in AI, AAAI Association for the Advancement of Artificial Intelligence, Nov 2022, Arlington, Virginia, United States. 3 p |
| Popis: | International audience; Thinking Fast and Slow (Kahneman 2011) provides a simple mental model of how human intelligence builds on components with complementing responsibilities and capabilities. In computer science in general, and artificial intelligence research in particular, these ideas are used to inspire new methods and architectures. We argue that many of those methods use the concept of Thinking Fast and Slow as a token reference, while not living up to the definitions of dual-process systems from psychology. For instance, 'fast' is seen as synonymous with neural, and the autonomy of 'fast' seen in, e.g., fixed social interactions of social agents is lost. We further highlight that these ideas are misused in saying that humans are flawed and AI systems can fix that. Given that human bias is highly context-dependent, such simplistic applications of dual-process theory to AI are likely to fail. Thus, the narrative that AI systems will provide users with rationality is flawed. In a work in progress, we survey and categorise (mis)use of prospect theory and other dual-process theories. Building AI systems on the ideas of Tversky and Kahneman is a step in the right direction for a more human-centered artificial intelligence. With this work we want to emphasise the many things to consider in building systems that are Thinking Fast and Slow, and that the community is only scratching that surface. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|