Unverified50% confidenceConceptExact time
Academia calls the phenomenon in which the decision-making process of recommender systems is difficult to explain the 'algorithmic explainability' problem, and distinguishes two levels: 'transparency' (disclosing how the system works) and 'explainability' (giving human-understandable explanations for specific recommendation results).
1
Sources
50%
Confidence
Long-term
Relevance
10/13/2026
First Seen
Sources
Related Entities
Related Claims
Unverified该项目引入主动消歧机制：当不确定性超过阈值时智能体不直接输出，而是主动向用户提出澄清问题68% similarUnverifiedAlgorithmic paternalism uses data and efficiency as a shield, wrapping interventions in the neutral guise of "system recommendations," making it difficult for critics to rebut.67% similarUnverifiedScoring the final explanation produced by an agent and constraining the actual operations that generate the explanation are two different things: the former is post-hoc review while the latter is pre-emptive restriction.67% similarUnverified当研究过程被高度封装进AI系统内部时，若推导过程对人类是黑箱，结论的可靠性、可解释性和可信度都将大打折扣67% similarUnverifiedIn regulated fields such as healthcare, finance, and compliance, a simple model that can clearly explain why a recommendation is given is often more valuable for deployment than a black-box model with slightly higher accuracy but no explainability.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1020869API
curl https://kongchang.com/api/v1/knowledge/claims/1020869MCP
get_claim(id=1020869)