Designing Warning Messages for Detecting Biased Online Product Recommendations: An Empirical Investigation
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
The increasing adoption of product recommendation agents (PRAs) by e-commerce merchants makes it an important area of study for information systems researchers. PRAs are a type of Web personalization technology that provides individual consumers with product recommendations based on their product-related needs and preferences expressed explicitly or implicitly. Whereas extant research mainly assumes that such recommendation technologies are designed to benefit consumers and focuses on the positive impact of PRAs on consumers’ decision quality and decision effort, this study represents an early effort to examine PRAs that are designed to produce their recommendations on the basis of benefiting e-commerce merchants (rather than benefiting consumers) and to investigate how the availability and the design of warning messages (a potential detection support mechanism) can enhance consumers’ performance in detecting such biased PRAs. Drawing on signal detection theory, the literature on warning messages, and the literature on message framing, we identified two content design characteristics of warning messages—the inclusion of risk-handling advice and the framing of risk-handling advice—and investigated how they influence consumers’ detection performance. The results of an online experiment reveal that a simple warning message without accompanying advice on how to detect bias is a double-edged sword, because it increases correct detection of biased PRAs (hits) at the cost of increased incorrect detection (false alarms). By contrast, including in warning messages risk-handling advice about how to check for bias (particularly when the advice is framed to emphasize the loss from not following the advice) increases correct detection and, more importantly, also decreases incorrect detection. The patterns of findings are in line with the predictions of signal detection theory. With an enriched understanding of how the availability and the content design of warning messages can assist consumers in the context of PRA-assisted online shopping, the results of this study serve as a basis for future theoretical development and yield valuable insights that can guide practice and the design of effective warning messages.
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.
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 | 0.012 | 0.003 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.001 | 0.008 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| 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 it