| 1 |
Interpretive blindness and the impossibility of learning from testimony
|
|
|
|
In: Dialogue and Discourse ; Proceedings of the 20th International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2021), ; International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2021) ; https://hal.archives-ouvertes.fr/hal-03138957 ; International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2021), International Foundation for Autonomous Agents and Multiagent Systems, May 2021, London (virtuel), United Kingdom (2021)
|
|
Abstract:
International audience ; We model interpretive blindness, a type of epistemic bias that posesa problem for learning from testimony, in which one acquires infor-mation from text or conversation but lacks direct access to groundtruth. Interpretive blindness arises when a co-dependence betweenbackground beliefs and interpretation leads to a dynamic processof bias hardening that impedes or precludes learning. We arguethat when bodies of data areargumentatively complete, even con-straints from hierarchical Bayesian learning designed to promotegood epistemic practices will fail to stop interpretive blindness.
|
|
Keyword:
[INFO.INFO-AI]Computer Science [cs]/Artificial Intelligence [cs.AI]; [SCCO.LING]Cognitive science/Linguistics; [SCCO]Cognitive science; Bayesian learning; formal epistemology
|
|
URL: https://hal.archives-ouvertes.fr/hal-03138957
|
|
BASE
|
|
Hide details
|
|
|
|
|
