Large language models reflect political bias in their generated knowledge
From Echo Chambers to Epistemic Monoculture: Large Language Models Present Temporally Contingent Partisan Alignments as Knowledge
Computation and Language
Summary
Large language models (LLMs) like Llama 3.1 generate text based on patterns in their training data, but they do not simply present facts neutrally. The authors show that these models carry political biases that match partisan identities, encoded in their internal structure. They also found that the models treat politically contested information as knowledge without distinguishing opinion from fact. This means that instead of helping people see a full range of views, LLMs may reinforce cultural and political divides by amplifying one-sided perspectives.
large language modelspartisan biastraining datapolitical polarizationecho chambersepistemic monocultureLlama 3.1knowledge framingalignment traininginformation environment
Authors
Wend K. Tam
Abstract
Large language models (LLMs) are rapidly becoming an interface between citizens and political information. They are often regarded as "a better Google." While this analogy might work for some instances, it is unintuitively problematic for democratic politics. A search engine retrieves human-authored documents, while a language model generates novel text that necessarily embeds invisible framing decisions. Because conveying knowledge involves framing, a system that generates answers cannot serve as a neutral conduit to "all human knowledge." Instead, these systems are becoming a new kind of political intermediary. Mechanistic evidence shows that partisan identity is encoded as a locatable geometric direction inside the Llama 3.1 8B model, and that alignment training masks rather than removes this structure. Building on that evidence, we present steering experiments that exploit a model's training cutoff in 2024. This cutpoint auspiciously falls just before a dramatic realignment in American politics marked by the second Trump administration and the MAHA transformation of health politics, providing us with a natural experiment. We find that the model presents temporally contingent partisan alignments as knowledge, with no mechanism for distinguishing fact from opinion. This reality moves the information environment beyond the echo chamber toward an epistemic monoculture where language models, purporting to summarize "all human knowledge" are, in actuality, simply magnifying the cultural and partisan divides inherent in their training data.