Unverified50% confidenceOpinionExact time
Opponents point out that if the training data comes from historical law enforcement records, the algorithm may entrench existing biases while cloaking itself in an appearance of 'objectivity.'
1
Sources
50%
Confidence
Long-term
Relevance
10/27/2026
First Seen
Sources
Related Entities
Related Claims
Unverified如果LoRA权重被逆向分析，攻击者可能从中推断出训练所用的私有数据内容，即记忆提取问题72% similarUnverifiedIf a model learns with police stop records over the years as labels, its output is essentially a reproduction of existing law enforcement behavior rather than an objective prediction of unlawful behavior.70% similarUnverifiedAn adversarial data stream means that the data are generated by a malicious or strategic adversary who can observe the algorithm's past behavior and construct inputs that make the algorithm err; statistical methods that rely on randomness assumptions may be fooled.68% similarUnverifiedAutomated screening algorithms may learn implicit biases from historical data, thereby violating fair housing law without any malicious intent.67% similarUnverifiedUnder the one-sided condition, when the historical data are less reliable than expected, the algorithm may suffer linear regret, which suggests that practitioners need to carefully assess the credibility of historical data.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1104470API
curl https://kongchang.com/api/v1/knowledge/claims/1104470MCP
get_claim(id=1104470)