Reducing the incidence of biased algorithmic decisions through feature importance transparency: an empirical study
Bibliographic record
Abstract
As firms move towards data-driven decision-making using algorithmic systems, concerns are raised regarding the lack of transparency of these systems which could have ramifications related to users’ trust and the potential for provoking discriminatory decisions. Although previous research has developed methods to improve algorithmic transparency, little empirical evidence exists regarding the extent of the effectiveness of these approaches. Drawing upon Rest’s theory of ethical decision-making and the literature on algorithmic transparency and bias, we investigate the effectiveness of feature importance (FI), a common transparency-enhancing approach, which illustrates the nature and the weights of the features utilised by an algorithm. Through an online experiment employing a fictitious tool that provided recommendations for selecting employees for a promotion-related training programme, we find that FI is effective when biased recommendations include direct discrimination (i.e. when individuals are treated less favourably on protected grounds such as gender); but is of little assistance when discrimination is indirect (i.e. when a criterion or practice that is apparently neutral, disadvantages a group of individuals who are of a protected class). Additionally, we propose a new transparency approach, using aggregated demographic information, to accompany FI in indirect discrimination circumstances and report the results of testing its effects.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.030 | 0.229 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.002 |
| Scholarly communication | 0.002 | 0.004 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.002 | 0.004 |
| Insufficient payload (model declined to judge) | 0.003 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".