Fairness in Markov decision processes

Fairness is a desirable property of decision   rules applied to a population of individuals. For example,   college admissions should be decided on variables that inform   about merit, but fairness may also require taking into account   the fact that certain communities are inherently disadvantaged.   At the same time, a person should not feel that another in a   similar situation obtained an unfair advantage. All this must be   taken into account while still caring about optimizing for a   decision maker's utility function. As another example, consider   mortgage decisions: while lenders should take into account the   creditworthiness of individuals in order to make a profit,   society must ensure that they do not unduly discriminate against   socially vulnerable groups. The problem becomes even more   challenging when we take into account potential uncertainties in   decision making models, which can make some notions of fairness   impossible to satisfy. This project will examine fairness in   decision making for a topic of the student's choice, but the   focus should always be in data driven problems.

The project can be either theoretical or practical.
  For the former, a solid background in mathematics is necessary,
  as the student will attempt to develop and prove fairness
  properties of algorithms. For the latter, the student should be
  able to envisage an appropriate case study, such as mortgage
  decisions or university admissions, collect data for it.
  Mathematical maturity is still necessary, but the emphasis is
  more on research methodology and programming skills. For this
  particular project, our emphasis is on the sequential aspects of
  fairness, more specifically fairness in Markov decision
  processes. This includes the tasks:

  • Formalize the decision problem.
  • Specify the fairness criteria used.
  • Develop an algorithm solving the problem while satisfying the fairness criteria.

 Skills. The student should have good working knowledge of
 calculus, probability and algorithms.

Benefits. The student will obtain background in
statistics, machine learning and constrained optimisation, as
well the links between these areas, as well as practical
experience in programming (mainly python/octave with some
performance critical functions in C) and/or optimality proofs.
Good theoretical or experimental results will lead to writing and
submitting a paper to a suitable peer-reviewed


Emneord: reinforcement learning, fairness, Markov decision processes
Publisert 25. sep. 2019 16:29 - Sist endret 25. sep. 2019 16:29


Omfang (studiepoeng)