Contribution to the handbook
Supporting · Assessed 2026-10-28
The packet supplies concrete adaptation details and reports option-order sensitivity, which are useful when considering safeguards and tests. These are study-specific technical details rather than a governance design or demonstrated solution.
Does this challenge the goal?
In-scope gap
Model adaptation and option-order sensitivity can inform testing of a possible citizen-preference estimation method within the accepted goal. The question does not challenge the goal itself.
The evidence describes one research setup and does not establish how those methods would perform in a governance system.
Why this category?
- Technical detail relevant to testing model outputs
- Reported option-order sensitivity and limits
The supplied evidence gives broad consistency results but not exact rates for each model. Reversing option order is described as an intended check, not a complete remedy for order effects.
This is an AI assessment of contribution to the goal, separate from whether the question has an answer.
Authority, accountability and resilience →
Tests, stop conditions and unresolved work →
Read the exploration goalNew evidence is being assessed. Any answer below is provisional; parts whose evidence changed or was withdrawn are withheld.
Describe the prompt inputs, treatment of proposal order, and parameter-adaptation method.
What we can say so far
The researchers used participant details and recorded choices as prompt context. The reported characteristics included age, political ideology, urban or rural zone, education, sex, and city and state. They presented proposal pairs in both orders to account for sensitivity to which option appeared first. For parameter adaptation, they fine-tuned the models with LoRA, adjusting about 12% of parameters while keeping the remaining weights fixed.
What remains open
- The evidence does not establish that including both proposal orders fully corrected order sensitivity.
- The evidence does not specify how missing or more varied participant characteristics were handled.
Conditions and limits
- This is the method reported by the researchers in the supplied preprint text.
- Presenting both orders was intended to address order sensitivity; the evidence does not establish that it eliminated the effect.
- The reported participant characteristics were represented in categories; the evidence does not explain how missing or more varied responses were handled.
Objections to consider
- The source reports that reversing proposal order could change predictions, so including both orders should not be taken as proof that order effects were corrected.
- Reported consistency varied across models, and the supplied evidence gives no basis to conclude that the order-handling procedure eliminated sensitivity.
Evidence and related ideas
Supporting evidence, qualifications and context play different roles in this answer.
supports
Adapting the models The researchers build prompts from participants’ details and recorded choices, include both orders of each proposal pair to address order sensitivity, and adapt about 12% of model parameters with LoRA while keeping the other weights fixed.
Directly describes participant details and recorded choices as prompt context, both proposal orders, and LoRA adaptation of about 12% of parameters.
Large Language Models (LLMs) as Agents for Augmented Democracy
Read full source: Large Language Models (LLMs) as Agents for Augmented Democracy (opens in a new tab) ↗