1. Photo by Bart van Overbeke. August 30, 2018
Diversity and Inclusion in
a CS classroom
Alexander Serebrenik
@aserebrenik
a.serebrenik@tue.nl
I would like to thank Vadim Zaytsev for inviting me and giving me this opportunity to talk to you about Diversity and Inclusion in a CS classroom.
I am particularly interested in understanding and supporting diversity of software development teams, their communication and collaboration. Indeed, software
development is essentially a collaborative activity similar to the harvest on the picture: people of di
ff
erent genders jointly work towards a shared goal.
2. First of all I would like to say that the
fi
eld of diversity and inclusion studies is broad and diverse, each one can select the subject they like, and of course I cannot cover
the
fi
eld in its entirety. There are many diversity axes, many educational settings and ultimately there is no silver bullet
3. Credits: Hridoy Khana; Bernard Vaillant, after Jusepe de Ribera (called Lo Spagnoletto)
So why are we talking about diversity in the
fi
rst place? There are several reasons for it.
The
fi
rst - and the most obvious - reason is that there are students of many di
ff
erent backgrounds in our classrooms and we need to
fi
nd ways to let them feel included.
The second reason is that by training more diverse developers force we can help the industry to address the needs of more diverse clients, since the developers should
be re
fl
ection of the customers
The third reason is attracting people who are not there, and whose talent potential we are missing.
4. And what is also crucial to understand is that diverse software engineering teams tend to work better, so we need to train diverse workforce.
5. tenure
gender
tenure
diversity
gender
diversity
Individual Team Process
turnover
productivity
comm. smell:
black cloud
Product
code smell:
long method
@aserebrenik
geography geographic
diversity
But what do we mean by saying that “diverse teams work better”? What does “diverse” mean?
When we are talking about diversity and inclusion, we usually talk about four kinds of variables
* individual, team, process, product
Diversity of course can only be measured at the team level
So what is diversity? Diversity is about what makes each of us unique and includes our backgrounds, personality, life experiences and beliefs, all of the things that make
us who we are. It is a combination of our di
ff
erences that shape our view of the world, our perspective and our approaches [1]. Ultimately diversity highlights di
ff
erences
and similarities: we see cars of di
ff
erent sizes, di
ff
erent colours, di
ff
erent functions, new ones and old ones, but ultimately all of them are cars.
[1] Only skin deep? Re-examining the business case for diversity, Deloitte 2011 https://www.ced.org/pdf/Deloitte_-_Only_Skin_Deep.pdf
6. Diversity in
Teams:
Bad or Good?
A priori, it is not clear whether diversity should be good or bad.
BAD: People prefer working with others similar to them in terms of values, beliefs, and attitudes. People categorise themselves into speci
fi
c groups. Members of own
group are treated better than outsiders. This suggests that communication should be easier in non-diverse teams.
GOOD: Diverse problem solvers outperform high ability problem solvers. Multicultural social networks promote creativity.
7. Bogdan Vasilescu, Daryl Posnett, Baishakhi Ray, Mark G. J. van den Brand, Alexander Serebrenik, Premkumar T. Devanbu,
Vladimir Filkov: Gender and Tenure Diversity in GitHub Teams. CHI 2015: 3789-3798
23K projects (671K devs, 10.7M commits)
8. Bogdan Vasilescu, Daryl Posnett, Baishakhi Ray, Mark G. J. van den Brand, Alexander Serebrenik, Premkumar T. Devanbu,
Vladimir Filkov: Gender and Tenure Diversity in GitHub Teams. CHI 2015: 3789-3798
We see hence that diversity, and in particular, gender diversity is good for productivity and does not a
ff
ect turnover. This means that we should expect that gender
diversity is embraced and women are welcomed.
9. Diversity in
Teams:
Bad or Good?
OK, now we know that more gender-diverse teams are more productive. However, you might remember that one of the main advantages of lack of diversity is an easier
communication. This is why in a follow up study we have focused on communication between software developers.
10. Qwerty1234qwer, CC BY-SA 4.0 <https://creativecommons.org/licenses/by-sa/4.0>, via Wikimedia Commons
Speci
fi
cally, we have looked at examples of broken communication. For example, one of the developers might be working in isolation and not communicating with their
peers; or there are two siloed subgroups that do not communicate, except through one or two of their respective members. Damian Tamburri and his co-authors have
called this kind of suboptimal communication patterns “community smells”. They have identi
fi
ed several such smells; in this talk I will focus on four.
11. Damian A. Tamburri, Philippe Kruchten, Patricia Lago, Hans van Vliet:
Social debt in software engineering: insights from industry. J. Internet Serv. Appl. 6(1): 10:1-10:17 (2015)
Organizational Silo: siloed areas of the developer community that essentially do not communicate, except through one or two of their respective members;
Lone-wolf: unsanctioned or de
fi
ant contributors who carry out their work irrespective or regardless of their peers, their decisions and communication;
Radio-silence: an instance of the “unique boundary spanner” [62] problem from social networks analysis: one member interposes herself into every formal interaction
across two or more subcommunities with little or no
fl
exibility to introduce other parallel channels;
Black-cloud: information overload due to lack of structured communication or cooperation governance;
12. Fabio Palomba, Damian Andrew Tamburri, Francesca Arcelli Fontana, Rocco Oliveto, Andy Zaidman, Alexander Serebrenik:
Beyond Technical Aspects: How Do Community Smells In
fl
uence the Intensity of Code Smells? IEEE Trans. Software Eng. 47(1): 108-129 (2021)
And indeed, these community smells have shown to be related to code smells, i.e., suboptimal organisation of communication between developers corresponds to
suboptimal organisation of the source code. For example, information overload represented by the black cloud co-occurs with lengthy methods and lone wolves with ill-
structured spaghetti code.
digraph smells {
rankdir=LR
BC -> LM [label="+"]
OS -> MC [label="+"]
OS -> LM [label="+"]
OS -> B [label="+"]
RS -> B [label="+"]
LW -> SC [label="+"]
LW -> FE [label="+"]
BC [style=
fi
lled,
fi
llcolor=gray, label="Black Cloud"]
OS [style=
fi
lled,
fi
llcolor=gray, label="Organizational Silo"]
RS [style=
fi
lled,
fi
llcolor=gray, label="Radio Silence"]
LW [style=
fi
lled,
fi
llcolor=gray, label="Lone Wolf"]
MC [style=
fi
lled, label="Misplaced Class"]
LM [style=
fi
lled, label="Long Method"]
B [style=
fi
lled, label="Blob"]
13. SC [style=
fi
lled, label="Spaghetti Code"]
FE [style=
fi
lled, label="Feature Envy"]
{rank=same;B;SC;FE;MC;LM}
{rank=same;LW;RS;OS;BC}
edge[style=invis];
OS -> BC -> RS -> LW;
}
Two last lines enforce the same order of the nodes as on the next image
14. So, we have decided to study how diversity is related to community smells.
15. Hofstede’s 6-D framework
In our recent study we have considered three kinds of diversity in GitHub open source teams: gender, geographic distances and national culture (Hofstede). Geographical
Dispersion de
fi
ned as the standard deviation of the set of physical distances between each community member—calculated using the origin country coordinates.
Power Distance Index. PDI expresses the degree to which the less powerful members of a society accept and expect that power is distributed unequally. People in
societies exhibiting a high level of Power Distance accept a hierarchical order in which everybody has a determined place. On the contrary, in societies with low Power
Distance, people strive to equalize the distribution of power and demand justi
fi
cation for power inequalities [35, 37].
Individualism vs. Collectivism. IDV represents the degree to which people in a society are integrated into groups. A high level of such a dimension indicates a society in
which individuals are expected to take care of only themselves and their immediate families. Conversely, a low level indicates a preference for a tightly-knit framework in
society in which individuals can expect their relatives or members of a particular group to look after them in exchange for unquestioning loyalty [35, 37].
Masculinity vs. Femininity. MAS represents a contrast between two preferences. The Masculinity side (high level) is de
fi
ned as “a preference in society for achievement,
heroism, assertiveness and material rewards for success”. In contrast, the Femininity side (low level) represents “a preference for cooperation, caring
for the weak and quality of life” [35, 37].
Uncertainty Avoidance. UAI expresses the degree to which the members of a society feel uncomfortable with uncertainty and ambiguity. Countries exhibiting high level
of UAI maintain rigid codes of belief and behavior and are intolerant of unorthodox behavior and ideas. Conversely, a low level of UAI indicates
societies that maintain a more relaxed attitude in which practice counts more than principles [35, 37].
Long vs. Short Term Orientation. LTO measures how much people are oriented toward a long-term outlook in contrast to a more short-term. A high degree in this index
16. (Long-Term) indicates that people encourage thrift and e
ff
orts in modern education as a way to prepare for the future. On the contrary, a lower
degree of this index (Short-Term) indicates that people tend to honor traditions and value steadfastness [35, 37].
Indulgence vs. Restraint. IVR refers to the degree of freedom that societal norms give citizens to ful
fi
ll their human desires. A high level (Indulgence) indicates a society
that allows relatively free grati
fi
cation of basic and natural human desires related to enjoying life and having fun. Conversely, a low level (Restraint) indicates a society that
controls grati
fi
cation of needs and regulates it using strict social norms [35, 37].
17. For the sake of simplicity, I am showing only three dimensions: power distance, individualism vs collectivism and masculinity vs femininity on three countries: China, the
Netherlands and the US.
We see that in Chinese culture large power di
ff
erences are more accepted than in other two country cultures, that US is the most individualistic culture among the three
and that the Duct culture tends to favour collaboration as opposed to competition.
18. Catolino, Palomba, Tamburri, Serebrenik, Ferrucci. Gender Diversity and Women in Software Teams: How Do They A
ff
ect Community Smells? ICSE SEIS, 2019,
pp. 11-20
Lambiase, Catolino, Tamburri, Serebrenik, Palomba, Ferrucci. Good Fences Make Good Neighbours? On the Impact of Cultural and Geographical Dispersion on
Community Smells, ICSE-SEIS, 2022
We are not going to look at all possible connections but we already see that at the very least on this particular dataset we could only
fi
nd relations between power
distance diversity and individualism diversity, on the one hand and community smells, on the other hand. First of all, this means that we could not see a relation between
such diversity variables as uncertainty avoidance diversity, masculinity/femininity diversity, etc and community smells.
digraph smells {
rankdir=LR
BC -> LM [label="+"]
OS -> MC [label="+"]
OS -> LM [label="+"]
OS -> B [label="+"]
RS -> B [label="+"]
GDI -> RS [label="+"]
GDI -> BC [label="-"]
GD -> RS [label="-"]
GD -> BC [label="-"]
GD -> OS [label="-"]
IDVD -> LW [label="-"]
IDVD -> RS [label="-"]
PDID -> RS [label="+"]
19. LW -> SC [label="+"]
LW -> FE [label="+"]
BC [style=
fi
lled,
fi
llcolor=gray, label="Black Cloud"]
OS [style=
fi
lled,
fi
llcolor=gray, label="Organizational Silo"]
RS [style=
fi
lled,
fi
llcolor=gray, label="Radio Silence"]
LW [style=
fi
lled,
fi
llcolor=gray, label="Lone Wolf"]
MC [style=
fi
lled, label="Misplaced Class"]
LM [style=
fi
lled, label="Long Method"]
B [style=
fi
lled, label="Blob"]
SC [style=
fi
lled, label="Spaghetti Code"]
FE [style=
fi
lled, label="Feature Envy"]
GDI [label="Gender Diversity Index"]
GD [label="Geographic Diversity"]
IDVD [label="National Culture Diversity:n Individualism"]
PDID [label="National Culture Diversity:n Power Distance"]
{rank=same;B;SC;FE;MC;LM}
{rank=same;LW;RS;OS;BC}
{rank=same;GD;IDVD;GDI;PDID}
edge[style=invis];
OS -> BC -> RS -> LW;
}
20. Lambiase, Catolino, Tamburri, Serebrenik, Palomba, Ferrucci. Good Fences Make Good Neighbours? On the Impact of Cultural and Geographical Dispersion on
Community Smells, ICSE-SEIS, 2022
Here we see that more gender diverse teams are less likely to develop communication problems such as Black cloud (women are known to take mediating roles). In the
context of Black Cloud and Organizational Silo, geographically dispersed team members can in
fl
uence negatively the presence of the smells. A possible motivation is
that managing people physically arranged in di
ff
erent parts of the world means using speci
fi
c management tools and protocols for communication and collaboration,
e.g., Trello and Jira which “nudge” the way of working towards a rather narrow, more disciplined approach (reducing the noise associated with Black Cloud.
digraph smells {
rankdir=LR
BC -> LM [label="+"]
OS -> MC [label="+"]
OS -> LM [label="+"]
OS -> B [label="+"]
RS -> B [label="+"]
GDI -> RS [label="+"]
GDI -> BC [label="-", penwidth=3, fontname="times-bold"]
GD -> RS [label="-"]
GD -> BC [label="-", penwidth=3, fontname="times-bold"]
GD -> OS [label="-"]
IDVD -> LW [label="-"]
IDVD -> RS [label="-"]
21. PDID -> RS [label="+"]
LW -> SC [label="+"]
LW -> FE [label="+"]
BC [style=
fi
lled,
fi
llcolor=gray, label="Black Cloud"]
OS [style=
fi
lled,
fi
llcolor=gray, label="Organizational Silo"]
RS [style=
fi
lled,
fi
llcolor=gray, label="Radio Silence"]
LW [style=
fi
lled,
fi
llcolor=gray, label="Lone Wolf"]
MC [style=
fi
lled, label="Misplaced Class"]
LM [style=
fi
lled, label="Long Method"]
B [style=
fi
lled, label="Blob"]
SC [style=
fi
lled, label="Spaghetti Code"]
FE [style=
fi
lled, label="Feature Envy"]
GDI [label="Gender Diversity Index"]
GD [label="Geographic Diversity"]
IDVD [label="National Culture Diversity:n Individualism"]
PDID [label="National Culture Diversity:n Power Distance"]
{rank=same;B;SC;FE;MC;LM}
{rank=same;LW;RS;OS;BC}
{rank=same;GD;IDVD;GDI;PDID}
edge[style=invis];
OS -> BC -> RS -> LW;
}
22. So what can we do to create a more inclusive environment?
23. People Materials
Groups
While there might be multiple ways of encouraging diverse students, we focus on three topics: people, teaching materials and organisation of the group projects.
24. Let us start with people. The
fi
rst step is to convey the message that CS education is a place for students from minoritized groups. And here the role models play an
important role. In this context I would like to mention the wonderful e
ff
ort originating from Twente and led by Marieke Huisman and Mariëlle Stoelinga called “Alice and
Eve” that also has an exhibition and a series of videos about 30 illustrious women in computing.
25. But of course there are also other minority groups. Unfortunately there seem to be no Dutch websites in these spaces but as teachers we can always mention similar
websites for CS people with disabilities
https://www.washington.edu/accesscomputing/resources/choosecomputing/accesscomputing-pro
fi
les
queer scientists in CS https://500queerscientists.com/?s=&category=63&location=0
black computer scientists https://academicin
fl
uence.com/rankings/people/black-scholars/computer-scientists
26. Photo by Center for Teaching. Vanderbilt University. 2012
While role models might be around, it is important that there are people who are closer to students, similar to them and can show the next step in their career (rather than
a remote goal). To the end Marieke Huisman recommends engaging diverse TAs and in particular for
fi
rst year courses since this is the
fi
rst encounter of students with
computer science.
27. Photo by Jacobs School of Engineering, UC San Diego
However, the previous discussion of role models is important for any engineering
fi
eld and maybe any scienti
fi
c
fi
eld (even if the minoritized groups might be di
ff
erent).
But is there anything special for CS? This is where teaching materials play a role.
29. Abraham Silberschatz, Henry F. Korth, S. Sudarshan. Database System Concepts. 6th edition.
Names should never be used as primary keys since people can change names for a multitude of reasons. Amy Ko who has recently changed her name spend more
than 100 hours and ca 2000 USD to change the names on multiple documents.
30. https://slideplayer.com/slide/3829388/
Do you remember the stable marriages problem? In the story as it was taught in the classic books, women and men each had preferences for partners: women ranked
men and men ranked women. The problem was to
fi
nd marriages that are stable, i.e., there is no pair of a woman and a man who would both rather have each other than
their current partners. Already the formulation of this problem reenforces gender binary and heteronormativity assuming that women and men can only be attracted to
each other. Moreover, taking a closer look at the algorithm we can conclude that it reinforces the stereotype of men as pursuing women and women as selecting the
suitors. In fact this is a problem on a bipartite graph, and it can be equally well presented as a story about job applicants and companies with positions they are applying
for.
31. Abraham Silberschatz, Henry F. Korth, S. Sudarshan. Database System Concepts. 7th edition.
This is yet another example: not only is this heteronormative (the seventh edition of the book was published in 2019) but it also uses typical Western names that do not
necessarily appeal to students from non-Western background. In this case one could have easily replaced names with generics like “two persons are married to each
other” etc
32. And of course, the group assignments are an inherent part of the curriculum and di
ff
erent students might experience working together with other students in very di
ff
erent
ways.
33. Yeray Barrios Fleitas (UT) Sterre van Breukelen (TUe) Gemma Catolino (JADS) Tom van Dijk (UT)
Xi Long (TUe) Alexander Serebrenik (TUe) Irvine Verio (UT) Andy Zaidman (TUD)
This is why supported by 4TU Diversity-Equity-Inclusion Funding we have conducted a study of gender and nationality composition in student teams, and you are the
fi
rst ones to see some of our results. The work is still in progress so some of our
fi
ndings might need to be re
fi
ned.
35. https://pxhere.com/en/photographer/767067
60% 93%
Our students have quite some experience working with diverse teams and more so with nationality than with gender: 60% have worked with gender-diverse teams vs
93% with nationality; most experiences with operating within such a diverse team are positive (orange) or very positive (red).
36. Photo by mohamed hassan form PxHere
We have observed no statistically signi
fi
cant di
ff
erences in the ways the project is experienced between respondents of di
ff
erent genders, those who have participated in
gender-diverse teams vs those who have not participated in such teams and idem for nationality-diverse teams. Most students are neutral or positive about their
experience.
39. 35
https://mybusinessnews.site/2020/06/22/computer-repair-is-important-for-any-business-or-individual/
https://www.businessinsider.com/how-to-teach-yourself-code-and-land-6-
f
igure-job-2019-7?r=US&IR=T
More tech
tasks (10)
More non-tech
tasks (21)
Men assumed women are not
to be trusted with coding so
they had to check every single
line they wrote (Woman)
No difference (63)
One of the participants has reported that men assumed women are not to be trusted with coding so they had to check every single line they wrote. This is a well-known
Prove-It-Again gender bias, known from the work of Joan Williams and Rachel Dempsey. This bias means that women are often measured at a stricter standard than man
and subsequently they must provide more evidence than men to demonstrate competence. This is an important observation since Imtiaz et al. did not observe this bias
in their study of professional development at GitHub.
Joan C Williams and Rachel Dempsey. What works for women at work, 2014.
Imtiaz et al. Investigating the E
ff
ects of Gender Bias on GitHub ICSE 2019
40. Next we take a look at the how Dutch vs non-Dutch students work.
41. No difference (76)
And again for many respondents there is no di
ff
erence between Dutch and international students.
42. No difference (76)
9-17 (31) 17-9 (35)
When it comes to di
ff
erences, they are related to traditional working hours: Dutch students tend to work during the o
ffi
ce hours (31 as opposed to 1 working outside the
o
ffi
ce hours); International students tend to work outside the working hours (35 as opposed to 5 working during the o
ffi
ce hours).
43. No difference (76)
9-17 (31) 17-9 (35)
Felt excluded (10)
The most problematic observation was that 10 respondents felt excluded…
44. No difference (78)
English
Talked more
More direct
Less direct
More open minded
More interesting di
ff
erences emerged when we look at the communication. We see that the Dutch students clearly show the attributes of the dominant group: they are
perceived as talking more and being more direct, while the international students appear less direct and having more troubles with English then their Dutch peers.
45. So the next question is of course: what can we do?
* engage in a conversation with students about importance of diversity and inclusion, in general. And of course, as Belgian, this is my favourite kind of diversity :)
* and speci
fi
cally about possible pitfalls, such as the members of the dominant group dominating discussion or women being required to repeatedly prove their
competence or being pushed out to non-technical tasks
46. To summarise…
First of all diversity in general is good - see higher productivity and lower number of community smells -and it is our duty to train diverse workforce.
47. To summarise…
Second to train such a diverse workforce we can consider three directions: people, such as famous computer scientists from minoritized groups and diverse TA; teaching
materials, in particular when they discuss humans or model human relations or data; and
fi
nally organisation of the group teams.
48. Together we can
create everything!
Alexander Serebrenik
@aserebrenik
a.serebrenik@tue.nl
because together we can create everything!