Yutaro AKITA

Bio

Ph.D. candidate in Economics at Penn State

ytrakita (at) gmail (dot) com

Fields: Decision theory · Microeconomic theory

cv arXiv GitHub Google Scholar

Working papers

Randomization and ambiguity perception

w/ Kensei Nakamura | September 2025

We provide a model of preferences over lotteries of acts in which a decision maker behaves as if optimally filtering her ambiguity perception. She has a set of plausible ambiguity perceptions and a cost function over them, and chooses multiple priors to maximize the minimum expected utility minus the cost. We characterize the model by axioms on attitude toward randomization and its timing, uniquely identify the filtering cost from observable data, and conduct several comparatives. Our model can explain Machina's (2009) two paradoxes, which are incompatible with many standard ambiguity models.

PDF BIB DOI
```
@unpublished{AkitaNakamura2025,
  author = {Akita, Yutaro and Nakamura, Kensei},
  doi = {10.48550/arxiv.2509.05076},
  note = {arXiv:2509.05076},
  title = {Randomization and Ambiguity Perception},
  year = {2025},
}
```

A simple proof of the representation theorem for betweenness preferences

January 2025

This paper presents a simple proof of Dekel's (1986) representation theorem for betweenness preferences. The proof is based on the separation theorem.

PDF BIB DOI

@unpublished{Akita2025,
  author = {Akita, Yutaro},
  doi = {10.48550/arxiv.2405.11371},
  note = {arXiv:2405.11371},
  title = {A Simple Proof of the Representation Theorem for Betweenness Preferences},
  year = {2025},
}

Work in progress

Experiments, decision rules, and costly information acquisition

This paper examines the behavioral implications of costly information acquisition on information acquisition problems. The primitive of my model is a preference relation over pairs of a decision rule and an experiment, which I call strategies. I develop axiomatic foundations for a model of costly information acquisition by a Bayesian decision maker. She chooses strategies as if balancing the benefit and cost of information. I also characterize the special cases of the costly information acquisition model where (i) the payoff and the cost are additively separable; (ii) the payoff and the cost are additively separable and the cost is posterior separable.
Bayesian representations of preferences over experiments

I axiomatically characterize a class of preferences over Blackwell experiments that have Bayesian representations. In the representation, the decision maker behaves as if maximizing the value of information for some decision problem and prior. My key axiom, concatenation independence, reflects Bayesian decision makers' independence of mixing experiments as long as likelihood ratios are preserved. I establish the essential uniqueness of the decision problem for a given prior. I also show that choices over a simple class of experiments are enough to elicit the decision problem.