Anchoring Bias in AI: Ensuring Accuracy and Integrity in Research
Bibliographic record
Abstract
Researchers can use artificial intelligence tools to streamline and enhance many steps in the research process. AI can quickly and efficiently curate research questions, identify relevant studies, synthesize prior research and identify gaps, gather and analyze data, and compose research findings and conclusions. Despite AI’s impressive capabilities, its responses are not necessarily accurate, complete, or free from bias. This presentation explores the ethics of responsible use of AI in research through the lenses of accuracy, honesty, and anchoring bias. Anchoring bias refers to a person’s tendency to over-rely on initial pieces of information that they receive, potentially discounting or ignoring other information that could confirm or disconfirm the veracity of the initial information. To mitigate this bias, researchers should think critically about AI-generated outputs, rather than over-relying on the first information they receive. This presentation provides researchers with strategies to mitigate the anchoring effect when they use AI in various stages of research: understanding AI’s limitations, nurturing awareness of anchoring bias, asking critical thinking questions to evaluate the accuracy of AI outputs, and employing other strategies to verify the accuracy and reliability of AI outputs. While AI can be a valuable research tool, researchers should remain ethically responsible for the rigor of their research methods and the veracity of the findings they report.
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 | 0.628 | 0.804 |
| Meta-epidemiology (narrow) | 0.002 | 0.004 |
| Meta-epidemiology (broad) | 0.006 | 0.003 |
| Bibliometrics | 0.011 | 0.011 |
| Science and technology studies | 0.013 | 0.075 |
| Scholarly communication | 0.038 | 0.033 |
| Open science | 0.008 | 0.025 |
| Research integrity | 0.016 | 0.021 |
| Insufficient payload (model declined to judge) | 0.011 | 0.007 |
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; the direct Gemma label and the distilled Codex classifier agree on what is shown here.
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".