Abstract:One aspect of the algorithmic lens in theoretical computer science is a view
on other scientific disciplines that focuses on satisfactory solutions that
adhere to real-world constraints, as opposed to solutions that would be optimal
ignoring such constraints. The algorithmic lens has provided a unique and
important perspective on many academic fields, including molecular biology,
ecology, neuroscience, quantum physics, economics, and social science.
This thesis applies the algorithmic lens to Bayesian epistemology.
Traditional Bayesian epistemology provides a comprehensive framework for how an
individual's beliefs should evolve upon receiving new information. However,
these methods typically assume an exhaustive model of such information,
including the correlation structure between different pieces of evidence. In
reality, individuals might lack such an exhaustive model, while still needing
to form beliefs. Beyond such informational constraints, an individual may be
bounded by limited computation, or by limited communication with agents that
have access to information, or by the strategic behavior of such agents. Even
when these restrictions prevent the formation of a *perfectly* accurate belief,
arriving at a *reasonably* accurate belief remains crucial. In this thesis, we
establish fundamental possibility and impossibility results about belief
formation under a variety of restrictions, and lay the groundwork for further
exploration.
What problem does this paper attempt to address?
The core problem that this paper attempts to solve is: **How to rationally form and update beliefs under the constraints of the real - world (belief formation)**. Specifically, the author Eric Neyman applies the algorithmic perspective to Bayesian epistemology and explores how individuals adjust their beliefs according to new information under practical limitations such as computation, information, and communication.
### Specific problems include:
1. **Computational limitations**:
- In Bayesian inference, accurately calculating posterior probabilities is usually computationally infeasible because it requires summing or integrating over high - dimensional spaces.
- For example, in complex models, calculating the Nash equilibrium is a PPAD - hard problem, meaning that fully rational behavior may involve computationally intractable problems.
2. **Information limitations**:
- In reality, individuals may not be able to obtain all relevant information, or the information itself may be incomplete or noisy.
- For example, in weather forecasting, the observational data may be coarse - grained or inaccurate, leading to limited accuracy of the prediction model.
- The paper explores how to aggregate the forecasts of multiple experts in the case of incomplete information and derive a reasonable comprehensive forecast.
3. **Communication limitations**:
- There may be limitations in the information exchange between individuals, such as limited bandwidth or incomplete information transfer.
- For example, in some cases, individuals can only convey information through short messages, and these messages may not be sufficient to convey complete information.
4. **Limitations of strategic behavior**:
- When multiple experts or agents are involved, their behavior may be strategic, that is, they will provide information according to their own interests.
- For example, in market forecasting, experts may provide biased forecasts for their own interests. The paper explores how to design incentive mechanisms in this case to ensure that the information provided by experts is as accurate as possible.
### Main contributions of the paper:
- **Theoretical framework**: Established the possibility and impossibility results of belief formation under various constraints.
- **Methodology**: Proposed new algorithms and techniques for effective belief updating under computational, information, and communication constraints.
- **Application examples**: Demonstrated the practical application value of these methods through specific examples (such as weather forecasting, market forecasting, etc.).
In summary, this paper aims to provide a new research direction for Bayesian epistemology by introducing the "algorithmic perspective", especially on how to deal with uncertainty more effectively and form reasonable beliefs in the face of various constraints in the real world.