Bias in First-Year Engineering Student Peer Evaluations

Résumé fourni par la source

This Complete Research paper describes a study on race, gender, and self-bias in first year engineering student's team peer evaluations.Our institution runs a first year introduction to engineering course with approximately 200 students that uses team projects over the span of the semester.Each project has 2-5 students per team and incorporates peer and self evaluations into each student's individual project grades.The researchers began this study to observe how racial, gender, and self-bias impact these peer evaluations.Peer evaluations are often employed in instances of group work, particularly in the undergraduate setting.These peer evaluations can present important information regarding team dynamic and distribution of workload.However, this method of assessment is also susceptible to both explicit and implicit biases, specifically in regard to race, gender, and self-bias.After identifying possible biases in our peer evaluation procedure, the researchers plan to examine methods to mitigate these biases.For each project, students submitted peer evaluations of themselves and each of their team members.The peer evaluations required the students to split 100 points amongst all team members including themselves with an optional written rationale for scores.The 1725 peer evaluation scores collected by this study were double-key entered into a database.Participants were also asked to self-identify as one of 6 gender options and 8 race options.If participants selected multiple races, they were assigned to the less common one.Two generalized linear regression models (GLMs) were used, one to estimate self-bias within each race/gender group (i.e.whether students give higher scores to themselves), and one to estimate how members in each race/gender group scored members of other groups, excluding the self-scores.Model coefficients significantly different from zero at the p<0.05 level indicated differences between groups and therefore possible evidence of biases.There were 160 participants, all of whom identified as either male or female.Due to small numbers, participants were combined into 3 race categories (White, Asian, or 'Other') for a total of 6 race/gender groups.Results showed that the students were significantly more likely to give themselves a higher score than other students on average, even after accounting for race/gender.This self-bias was consistent in both genders.White males gave scores to Other males that were significantly lower than what they gave to all other groups and significantly lower than what Other males received from White females or Asian males.This suggests a possible negative bias from White males to Other males.Similarly, Asian males gave scores to Other males and Other females that were significantly higher than what they gave to all other groups and significantly higher that what Other males and Other females received from all other groups.This suggests a possible positive bias from Asian males to Other males and Other females.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.