- Using a dataset build a hiring system (predict “hire”, “rejected at interview”, or “rejected at pre-interview”)
- Deploying models that impact people requires careful ethical decision making.
- One of the components of ethical AI is Systematicity
- Arbitrariness: The inherent error in the model due to overfitting/underfitting or issues in the data (unpredictability, unconstrained, and unreasonable)
- Systematicity: A model (algorithm) with inherent arbitrariness used on a large scale within an entire sector (e.g. hiring, education, and loans) can produce systematic discrimination.
- When deploying talent models at scale, we have to design systems that have some degree of randomness in order to ensure that individuals are not arbitrarily blocked from economic opportunities
- Since we are unable to completely remove arbitrariness from a model, we may at least reduce its impact.
- The authors of the paper* outline two primary ways of addressing systematicity:
- Ensemble model method: Training a set of models (with similar accuracy) and randomly drawing from the set of models at prediction time (I chose to implement this one)
- Directly introducing (bounded) randomness to scores at prediction time
- A predict() function that implements one or more mechanisms for addressing systematicity
- A set of metrics and associated charts that allow someone to measure and evaluate the effectiveness of my systematicity mitigation strategy.
The following are the two desirable behaviors of an ensemble technique addressing systematicity:
- Fairness: ensures that every deserving candidate is at least ‘hired’ once by an ensemble model.
- Arbitrariness: ensures that all deserving candidates have an equal chance of being “hired” by our ensemble model.
*Creel, Kathleen and Hellman, Deborah, The Algorithmic Leviathan: Arbitrariness, Fairness, and Opportunity in Algorithmic Decision Making Systems (February 15, 2021). Virginia Public Law and Legal Theory Research Paper No. 2021-13, Available at SSRN: https://ssrn.com/abstract=3786377