The struggle to measure gender bias in remote programming environments
Links table
Abstract and 1 introduction
1.1 Twincode platform
1.2 Experimental studies
1.3 other sex identities and 1.4 paper structure
2 related work
3 original study (Seville December, 2021) and 3.1 participants
3.2 Implementation of the experiment
3.3 factors (independent variables)
3.4 Response variables (dependent variables)
3.5 Erbak variables
3.6 data analysis
4 The first reproduction (Berkeley May, 2022)
4.1 Participants
4.2 Exercise implementation
4.3 Data Analysis
5 Discussion and authority threats and 5.1 The activation of the cause building – treatment
5.2 Activating the structure of influence – standards
5.3 Samples from the population – the participants
6 conclusions and future work
6.1 The symmetrical copies in a different cultural background
6.2 Use Chatbots as partners and coding of the artificial intelligence speech
Data groups and compliance with moral standards, recognition and references
A. The Question Insights No. 1 and #2
for. TWINCODE user development
C. TAG-A-Chat user interface
5 Discussion and authority threats
In this section, the original study is discussed and external repetition. Since the main concerns revolve around their trial validity threats regarding employment and sampling, the discussion is organized on this type of threats, especially those that were not previously discussed in describing the similar copies changes in the 4.1 and 4.2 sections.
5.1 Carry Building – Treatment
Activating gender bias in treatment is not a trivial task, and according to the results obtained, we may not have designed our treatment as we aim, and thus threaten construction health.
Given our experimental design, telling people that they will cooperate with a more frank man or woman could have caused many of them to participate in noticing this fact, and they behave abnormally, and they may not have mentioned it unintentionally during the chatting of the chat, and thus discovered that they were deceiving the gender of their partner and adopting this study.
However, although the silhouettes in the original experience (see Figure 9 A) had an effective 60 % effectiveness (see Table 4), when they were changed in the symmetrical copies of what we thought they were more clear. Regardless of the change of deities, this decrease in the effectiveness of treatment was affected by other factors, such as the distant preparation, which increased the possibility of deviations compared to the environment under control such as the laboratory session, and the comment was also suspended in section 4.2.2. There can be other factors that are the decrease in the duration of the tasks within the husband and the second and third questionnaires, as previously discussed in the 4.2.3 section, and the so -called exhaustion of fatigue [49]That is, fatigue and exhaustion caused by the prolonged use of video conference platforms during the Covid -19 pandem [41, 54].
As was suspected in section 6.2, we evaluate the use of Chatbots with a design within the subjects in future symmetry copies to improve treatment and thus mitigate this threat to build health.
5.2 Activating the structure of influence – standards
The main goal of our work is to explore the effects of gender bias in the husband’s programming remotely. Because of this exploratory nature, we have applied systematic Trinity [13]Monitor the phenomenon is one of the largest possible number of views, with an activation based on 45 response variables of different types measured during a reasonable reaction period.
After saying this, during the coding of the chat words, some authors who are in the fifties at the moment of writing this article look at strong differences in how the youth generation [15]Communication compared to the way we did when we were of their age. With the entitled caution, taking into account the strong social and political environment in Spain and the United States against any kind of gender discrimination, we believe that it is possible that the presence of gender bias in people from our generation (the tenth generation) has decreased after two generations, although we have no adequate evidence to confirm this. In addition, if the bias between the sexes persist, it can be most of the subjects of self -censorship, which impedes the discovery of its effects. To improve this position, we are currently developing the Twentode platform to include more measures, and we also study the inclusion of quality research that may lead to new results in future reputation by expanding the collected information scope.
5.3 Samples from the population – the participants
5.3.1 low women in the original study
Unfortunately, the small percentage of women in STEM studies is a common issue in most higher education institutions [1, 51]. The decrease in the number of posts in the original study was an obstacle to studying whether the gender bias was mainly a male feature or if it was also present in women in any way. However, the percentage of women increased significantly in the first symmetry versions without important results about the interaction of the sex of the subject with other factors.
5.3.2 Small size of the sample in the symmetrical copies
It is assumed that the size of the smaller sample in the symmetrical copies and low effectiveness of treatment is a clear threat to the inferiority of the validity that can only be mitigated by taking temporary results and performing more similar copies with larger samples and alternative experimental designs in the future.
5.3.3 Use students as materials
Although in other experimental studies in which themes are software engineering students, the results can be circulated reasonably to a broader society because experimental tasks do not usually require high levels of industrial experience [43]And students, who are the next generation of professionals, are close to the population under study [19, 34, 45]The differences between generations in Section 5.2 and the lack of conclusive results that make this very difficult in our case.
Authors:
(1) Amador Duran, I3us Institute, Sevilla University, Seville, Spain and Al -Nawal Laboratory, Seville University, Seville, Spain ([email protected]);
(2) Pablo Fernandez, i3us Institute, Universidad De Sevilla, Sevilla, Spain and Score Lab, Universidad de Sevilla, Sevilla, Spain ([email protected]);
(3) Betriz Bernardez, i3us Institute, Universidad De Sevilla, Sevilla, Spain and Score Lab, Universidad de Sevilla, Sevilla, Spain ([email protected]);
(4) Nathaniel Winmann, Computer Science Department, University of California, Berkeley, Berkeley, USA ([email protected]);
(5) Aslinhan Akalin, Computer Science Department, University of California, Berkeley, Berkeley, USA ([email protected]);
(6) Armando Fox, Computer Science Department, University of California, Berkeley, Berkeley, USA ([email protected]).