Norsk English

Causal AI: A systematic review of state of the art and relevant systems

The goal of this thesis is to systematically review the current state of the art in the area of Causal AI, with a focus on causal inference in general, and causal representation learning in particular

Contact person

Keywords: Causality, Causal inference, AI, Causal representation learning

Causal AI includes different techniques, like causal graphs and simulation, that help uncover causal relationships to improve decision making. According to the 2022 Gartner Hype Cycle it will take 5 to 10 years for causal AI to reach mainstream adoption, the business benefits are expected to be high — enabling new ways of performing horizontal or vertical processes that will result in significantly increased revenue or cost savings for an enterprise. Causal AI benefits include: efficiencies from adding domain-knowledge to bootstrap causal AI models with smaller datasets; greater decision augmentation and autonomy in AI systems; more robustness and adaptability by leveraging causal relationships that remain valid in changing environments; Better explainability by capturing easy-to-interpret cause-and-effect relationships.

The foundations of Causal AI were established in the area of causal inference - an interdisciplinary field that has seen important contributions have computer science, econometrics, epidemiology, philosophy, statistics, and other disciplines. Causal inference is the process of determining the independent, actual effect of a particular phenomenon that is a component of a larger system. Empirical research on causal inference is devoted to identifying, estimating, testing, and evaluating causal relationships between socioeconomic phenomena; this typically involves the construction of directed graphs to represent these causal relationships. Unlike the traditional causal inference approach, which uses causal graphs to connect random variables to complete the causal discovery and reasoning hypothesis task, the problem of causal representation learning has recently attracted more attention. Causal representation learning refers to learning variables from data. It does not require prior knowledge of human partitioning to learn information from data. Directly defining objects or variables related to a causal model is equivalent to directly extracting a more detailed, coarse-grained model of the real world.

In this context, the goal of this thesis is to systematically review the current state of the art in the area of Causal AI, with a focus on causal inference in general, and causal representation learning in particular.

Work to be done:

  • Systematic state-of-the-art review in the area of Causal Ai.
  • Evaluate existing tools in the area of causal inference (e.g., doWhy, LiNGAM).

References: