| paper | d4-conf-ijcai-2003-0049 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | no |
The paper’s named results are representation theorems and correspondences between representation constraints and Savage’s postulates; none asserts an algorithm, a complexity bound, or hardness. Turning the single decision maker into a population of agents would require adding an aggregation rule and a collective objective, so it would create a different problem rather than continuize a result here.
fails bit a — no named computational result to mirror
None: the arbitrary-preference representation, its monotone and additive strengthening, the fixed-domain representation, and the Savage-postulate correspondence are representational results; the paper states no computational result.
I can’t make a qualifying case for a continuous mirror from this paper: it contains no named computational result to anchor one to. Its Section 3 theorem says every preference relation on acts has a generalized expected utility representation. Its Section 4 theorem relates selected constraints on the representation to the corresponding Savage postulates. Both are representation results; neither gives an algorithm, a complexity classification, or a computational problem with an encoded input.
There is a tempting population story in the paper’s motivation about software agents: let a type be a complete agent configuration, and let \(\mu_t\) be its share in a large fleet. But the paper studies one decision maker’s preferences over acts. Its probability measure describes uncertainty over states of the world, not a population of agents. To make the fleet story into a problem, we would need to add a collective decision task and an objective the paper does not supply. That would be a new problem, not a continuous mirror of a result here.
So there is no honest anchor-by-anchor problem to hand to a prover, and no tractable-versus-hard expectation supported by this paper. Its framework might be useful in later work on populations of software agents, but that is only a research lead. The weakness is decisive for this task: the paper offers no computational result whose population continuization could be assessed.
The proponent raised no anchor, so there is no continuous question to defeat one by one. That absence is substantive: the paper’s main results are representation theorems. One says every preference relation on acts has a generalized expected utility representation; the other connects constraints on those representations to Savage’s postulates. Neither states a computational problem, gives a finite input model, or establishes an algorithm or complexity bound.
The strongest rescue would be to model a large fleet of software agents as a distribution over complete agent types, then ask which act the fleet should choose or how to synthesize a common decision rule. But the paper supplies no rule for aggregating agents’ preferences, no shared choice objective, and no computational task for that population. Its plausibility measure describes one decision maker’s uncertainty over states of the world; treating it as a distribution over people changes what the model means. Adding aggregation and an objective could create a worthwhile computational social choice problem, but its computational content would come from those additions, not from a mirror of either theorem.
The universal claim that no such scenario could ever be worthwhile is too strong to prove. A later multi-agent extension might provide a sound basis for one. On this paper’s results, though, there is no population computation to continuize: the possible fleet story is a research lead, not a continuous mirror.
The adversarial triple: the proponent anchors on up to three named results; the opponent sees that case and must defeat every anchor; the judge decides which case convinced it. These are the pipeline’s own outputs, generated by tools/triple_run.py — no human edited them. The paper’s own text is not reproduced here beyond the quoted statement above.