paint-brush
How Twincode Uncovers Gender Bias in Remote Pair Programmingby@pairprogramming

How Twincode Uncovers Gender Bias in Remote Pair Programming

by Pair Programming TechnologySeptember 17th, 2024
Read on Terminal Reader
Read this story w/o Javascript
tldt arrow

Too Long; Didn't Read

The section details the variables used in the twincode study, including current and potential future metrics. It covers key considerations and abbreviations for each variable.
featured image - How Twincode Uncovers Gender Bias in Remote Pair Programming
Pair Programming Technology HackerNoon profile picture

Abstract and 1 Introduction

1.1 The twincode platform

1.2 Related Work

2 Research Questions

3 Variables

3.1 Independent Variables

3.2 Dependent Variables

3.3 Confounding Variables

4 Participants

5 Execution Plan and 5.1 Recruitment

5.2 Training and 5.3 Experiment Execution

5.4 Data Analysis

Acknowledgments and References

5.4 Data Analysis

During the manual tagging of the dialog messages, all pairs in which the gender of any of the peers is disclosed in any way are excluded from the data analysis. Then, before analyzing response variables, the internal consistency of the questionnaire data is checked using Cronbach’s alpha and Kaiser criterion. After that, for every dependent variable v, we compute the distance between the two in–pair tasks as the absolute value of the difference, i.e. | v(t2) – v(t1) |. Ideally, this distance should be lower for the students in the control group (no information about partners’ genders) than for those in the experimental group (with two different perceived partners’ genders at t1 and t2). Therefore, for every variable, we perform a t–test to detect distance differences between the groups.


Then, using the data from the experimental group only, we perform a t–test to detect differences in the scores of every dependent variable between perceived partner’s gender, i.e. to detect differences in the scores when partners are perceived as men vs. as women. Finally, to detect a potential interaction between the perceived partner’s gender and the subject’s gender, we perform a mixed–model ANOVA with the perceived gender as a within– subjects variable and subject’s gender as a between–subjects variable. As complementary analyses, we also study (i) the correlation between the induced and the perceived gender for the subjects in the experimental group, and the distribution of the perceived gender (if any) in the control group; and (ii) the potential cultural impact of the different locations at which the experiment is carried out.


All the data analysis will be performed using R scripts, that will be available in a public repository together with the datasets in the corresponding laboratory package.


Authors:

(1) Amador Durán, SCORE Lab, I3US Institute, Universidad de Sevilla, Sevilla, Spain ([email protected]);

(2) Pablo Fernández, SCORE Lab, I3US Institute, Universidad de Sevilla, Sevilla, Spain ([email protected]);

(3) Beatriz Bernárdez, I3US Institute, Universidad de Sevilla, Sevilla, Spain ([email protected]);

(4) Nathaniel Weinman, Computer Science Division, University of California, Berkeley, Berkeley, CA, USA ([email protected]);

(5) Aslı Akalın, Computer Science Division, University of California, Berkeley, Berkeley, CA, USA ([email protected]);

(6) Armando Fox, Computer Science Division, University of California, Berkeley, Berkeley, CA, USA ([email protected]).


This paper is available on arxiv under CC BY 4.0 DEED license.