People Perceive Algorithmic Assessments as Less Fair and Trustworthy Than Identical Human Assessments
Bibliographic record
Abstract
Algorithmic risk assessments are being deployed in an increasingly broad spectrum of domains including banking, medicine, and law enforcement. However, there is widespread concern about their fairness and trustworthiness, and people are also known to display algorithm aversion, preferring human assessments even when they are quantitatively worse. Thus, how does the framing of who made an assessment affect how people perceive its fairness? We investigate whether individual algorithmic assessments are perceived to be more or less accurate, fair, and interpretable than identical human assessments, and explore how these perceptions change when assessments are obviously biased against a subgroup. To this end, we conducted an online experiment that manipulated how biased risk assessments are in a loan repayment task, and reported the assessments as being made either by a statistical model or a human analyst. We find that predictions made by the model are consistently perceived as less fair and less interpretable than those made by the analyst despite being identical. Furthermore, biased predictive errors were more likely to widen this perception gap, with the algorithm being judged even more harshly for making a biased mistake. Our results illustrate that who makes risk assessments can influence perceptions of how acceptable those assessments are - even if they are identically accurate and identically biased against subgroups. Additional work is needed to determine whether and how decision aids should be presented to stakeholders so that the inherent fairness and interpretability of their recommendations, rather than their framing, determines how they are perceived.
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 distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
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 teacher head, 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".