Search papers, labs, and topics across Lattice.
Affiliation:
3
0
7
14
It is found that LLMs frequently adjust the readability of feedback to education levels when these are explicitly mentioned, and on the other hand, implicit conditions produce unpredictable biases, such as in question answering, where responses from lower-education levels receive lower sentiment scores.
Over a third of real-world information-seeking queries are high-risk, exposing critical failures in LLM responses, especially for analytical tasks.
LLMs struggle to mirror the diverse language styles of different socioeconomic groups, tending to amplify upper-class linguistic norms and potentially reinforcing social biases.