MétaCan
Menu
Back to cohort
Record W4386470914 · doi:10.1101/2023.09.03.556128

How prior and p-value heuristics are used when interpreting data

2023· preprint· en· W4386470914 on OpenAlexaff
E. Hermer, Ashley A. Irwin, Dominique G. Roche, Roslyn Dakin

Bibliographic record

VenuebioRxiv (Cold Spring Harbor Laboratory) · 2023
Typepreprint
Languageen
FieldComputer Science
TopicData Analysis with R
Canadian institutionsCarleton University
Fundersnot available
KeywordsHeuristicsHeuristicBayesian probabilityValue (mathematics)Null hypothesisStatistical hypothesis testingComputer sciencePsychologyTest (biology)Data scienceCognitive psychologyStatisticsMachine learningArtificial intelligenceMathematics

Abstract

fetched live from OpenAlex

ABSTRACT Scientific conclusions are based on the ways that researchers interpret data, a process that is shaped by psychological and cultural factors. When researchers use shortcuts known as heuristics to interpret data, it can sometimes lead to errors. To test the use of heuristics, we surveyed 623 researchers in biology and asked them to interpret scatterplots that showed ambiguous relationships, altering only the labels on the graphs. Our manipulations tested the use of two heuristics based on major statistical frameworks: (1) the strong prior heuristic, where a relationship is viewed as stronger if it is expected a priori , following Bayesian statistics, and (2) the p-value heuristic, where a relationship is viewed as stronger if it is associated with a small p-value, following null hypothesis statistical testing. Our results show that both the strong prior and p-value heuristics are common. Surprisingly, the strong prior heuristic was more prevalent among inexperienced researchers, whereas its effect was diminished among the most experienced biologists in our survey. By contrast, we find that p-values cause researchers at all levels to report that an ambiguous graph shows a strong result. Together, these results suggest that experience in the sciences may diminish a researcher’s Bayesian intuitions, while reinforcing the use of p-values as a shortcut for effect size. Reform to data science training in STEM could help reduce researchers’ reliance on error-prone heuristics. Significance Statement Scientific researchers must interpret data and statistical tests to draw conclusions. When researchers use shortcuts known as heuristics, it can sometimes lead to errors. To test how this occurs, we asked biologists to interpret graphs that showed an ambiguous relationship between two variables, and report whether the relationship was strong, weak, or absent. We altered features of the graph to test whether prior expectations or a statistic called the p-value could influence their interpretations. Our results indicate that both prior expectations and p-values can increase the probability that researchers will report that ambiguous data shows a strong result. These findings suggest that current training and research practices promote the use of error-prone shortcuts in decision-making.

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 imitation

Not 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.

metaresearch head score (Codex)0.002
metaresearch head score (Gemma)0.002
Version: codex-gemma-dda1882f352aValidation status: machine_predicted_unvalidated
Candidate categoriesMeta-epidemiology (narrow), Scholarly communication, Open science
Consensus categoriesOpen science
DomainCandidate signal: none · Consensus signal: none
Study designCandidate signal: Simulation or modeling · Consensus signal: none
GenreCandidate signal: Methods · Consensus signal: none
Teacher disagreement score0.729
Threshold uncertainty score0.999

Codex and Gemma teacher scores by category

CategoryCodexGemma
Metaresearch0.0020.002
Meta-epidemiology (narrow)0.0010.001
Meta-epidemiology (broad)0.0010.000
Bibliometrics0.0010.001
Science and technology studies0.0000.000
Scholarly communication0.0030.001
Open science0.0070.016
Research integrity0.0000.001
Insufficient payload (model declined to judge)0.0000.000

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.

Opus teacher head0.041
GPT teacher head0.250
Teacher spread0.209 · how far apart the two teachers sit on this one work
Validation statusscore_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it

Classification

machine, unvalidated

Machine predicted; both teacher heads agree on what is shown here.

Study designSimulation or modeling
Domainnot available
GenreMethods

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".

Quick stats

Citations0
Published2023
Admission routes1
Has abstractyes

Explore more

Same venuebioRxiv (Cold Spring Harbor Laboratory)Same topicData Analysis with RFrench-language works237,207