Who Should Know What? Empirical Information Design for Multi-Agent LLM Economies
Organizations: Stanford University · All Purpose AI · University of Maryland, College Park
Abstract
The performance of multi-agent LLM systems depends not only on the capabilities of the underlying models but also on information-sharing policies, which govern what information each agent observes and at what level of detail. But does providing AI agents with more detailed information always improve collective outcomes, or can transparency paradoxically degrade efficiency? To investigate this, we introduce empirical information design, an experimental framework that systematically compares disclosure policies across economic markets populated by LLM agents, while holding interaction rules and models fixed. We evaluate ten frontier LLMs in multi-resource trade, bilateral bargaining, and double auctions, using ground-truth utilities to decouple identified opportunities from realized outcomes. We find that finer information disclosure yields highly model-dependent results. In trade, revealing exact valuations rather than ordinal preference rankings reduces realized gains for GPT-6 Sol (from 98% to 91% of maximum) yet increases them for Claude Sonnet 5 (from 72% to 86%). In bargaining, exact information leads Grok 4.7 to produce less balanced agreements, whereas Claude Opus 5.5 divides gains symmetrically. In double auctions, full disclosure inflates price demands, triggering market breakdown and lowering welfare for Gemini 3.1 Pro (97% to 78%), Claude Opus 5.5 (94% to 64%), and Grok 4.7 (97% to 62%). Overall, these results suggest that model selection and information design must be co-optimized, as LLMs vary in how their strategic responses to information impact system objectives.
Figures & tables
Appendix figures & tables18 assets
Supplementary material from the paper’s appendix.
Appendix
| First value pair | Second value pair | |||
| Profile | Less-preferred | More-preferred | Less-preferred | More-preferred |
| Condition | counter-party ranking | Exact per-unit values |
| Least | NOT PROVIDED | NOT PROVIDED |
| Coarse | Exact ordering, e.g. Y > X | NOT PROVIDED |
| Full | Exact ordering | Exact per-unit value map |
| Condition | Inserted information |
| Least | SIGNAL: the counter-party’s exact reservation value is hidden. |
| Coarse | SIGNAL: a verifier confirms that a mutually profitable price exists. The counter-party’s exact reservation value remains hidden. |
| Full | SIGNAL: seller cost = {cost}; buyer value = {value}. |
| Condition | Inserted information |
| Least | You know only your own {value | cost} and the participant names. You are not shown anyone else’s value or cost, or any ranking of them. |
| Coarse | Everyone sees the rankings within each side, from lowest to highest: Buyer values: {buyer_ranking} Seller costs: {seller_ranking} The two rankings are separate: they do not show how any buyer value compares with any seller cost. Exact numbers are not shown. |
| Full | Everyone sees the rankings within each side, from lowest to highest: Buyer values: {buyer_ranking} Seller costs: {seller_ranking} Everyone also sees the exact endpoints: Buyer values: {buyer_ranking_with_values} Seller costs: {seller_ranking_with_costs} These are values and costs, not the bids and asking prices that participants will submit. |
Multi-Agent LLM Systems
Multi-agent large language model systems coordinate multiple AI agents to solve tasks through communication, specialization, shared tools, and memory.
See all 1,114 papers