Thread by @DiverseInAI

3rd workshop on Artificial Intelligence Diversity, Belonging, Equity, and Inclusion (AID BE I) at #AAAI: a livetweet thread by @banazir

#DiversityInAI #DiverseInAI
@RealAAAI @WiMLworkshop @black_in_ai @_LXAI @QueerinAI @AiDisability #IndigenousInAI
@wimlds @BlackWomenInAI

1/🧵

Welcoming remarks from @banazir:

There will again be a special issue of Proceedings of Machine Learning Research (#PMLR, an imprint of @JmlrOrg's Journal of #MachineLearning #research) on this workshop.

Video recordings of the workshop will be at DiverseInAI.org.

2/🧵

First presentation: "Hello* - A Beginner's Guide to the Conference Galaxy" - Bethany Chamberlain, Dovile Juodelyte and Veronika Cheplygina

@chamberlain_ba @DrVeronikaCH

Great use of @Mentimeter!

3/🧵

Questions by @banazir on #DataStorytelling, #infovis, & #machinelearning and by @skoularidou on imputation of missing data (esp. qualitative data) vs. semisupervised/unsupervised learning (esp. similarity-based learning) using survey data.

4/🧵

Second presentation: "Responsible Research and Innovation for a Trustworthy AI: A Brief Report of Recent activities in Ethics and Gender" by Francesca Alessandra Lisi (@f_a_lisi). Applications in #NLP, especially #chatbots, #DialogueAgents, & #ComputationalLaw. #complaw

5/🧵

Debate on #AIethics, #Law, and #Forensics. #CompLaw #argumentation

6/🧵

Workshop of @ACMwomENcourage in Rome by @unibait. "Gendering ICT"

7/🧵

From talk by @f_a_lisi: "Looking at AI through gender lens": gender is a paradigmatic facet, gender bias concerns more than half the world population; #UNAgenda2030; #TrustworthyAI not possible without #gender #inclusion, #accountability

@UN_Women

8/🧵

Questions by @skoularidou: gender is not strictly binary nor the only facet of representation - our data are very biased in terms of inclusion and fairness. How can we diversify our data intersectionally, and promote marginalized unity over division and formal "50%" equity?

9/🧵

First invited talk, by @arjunsubgraph, representing @QueerinAI leadership:

"Prioritizing Grassroots D&I Activism: Queer in AI"

Synopsis of Arjun's personal journey, QAI's activities including fundraising for queer student researchers and support of inclusion & equity.

10/🧵

Thanks to @skoularidou graciously yielding some of her speaking time, we have this talk within a talk by @arjunsubgraph: "Harms of Gender Exclusivity and Challenges in Non-Binary Representation in Language Technologies"

#DiversityInNLP #InclusionInAI #InclusionInNLP

11/🧵

Remark by @skoulidarou on #allyship: identification and allyship can be orthogonal! #GenderIdentity #InclusiveIdentity

12/🧵

Second invited talk, by Maria @skoularidou, founder of @AiDisability and Women in Data Science and Statistics (@in_statistics) - "AIDBEI@AAAI-2022: On Equality, Diversity, and Inclusivity"

13/🧵

Maria @skoularidou relates her personal experience at @NeurIPSConf 2019: observing access limitations and underrepresentation of persons with disabilities, and advocating for disability accomodation. She presents the following action points for @AiDisability and allyship:

14/🧵

Short break and impromptu "meet the {pets | kids}" session. We should do this again! 😊

15/🧵

Panel discussion on the near future of diversity in artificial intelligence: Perspectives, Education, and Outreach.

Panelists: @jeffadoctor @ShawnTsosie4 @skoularidou @QuickResolute @vuisnotabot
@RealAAAI #AAAI2022 #Panel #DiversityPanel #DEI #DiversityInAI #AAAI44

16/🧵

Q1 (RISKS, HARMS, TECHNO-OPTIMISM, AND CRITICAL PERSPECTIVES): We are in a period when the pandemic, climate change, wars, and other catastrophes have both highlighted and derailed certain aspects of AI's human-beneficial uses and abuses. (contd.)

17/🧵

Q1 (contd.)

Recently there has arisen a tension between techno-optimism and critique of AI's risks and harms. Can you address some of the ramifications with respect to diversity and inclusion in AI?

#TechnoOptimism #TechnoCriticism #DiversityInAI #DiversityInML #AAAI2022

18/🧵

A1 (@QuickResolute): we have seen a lot of issues around underrepresentation, fear of potential loss of work due to AI/ML development, bias in ML systems, and representation (not just within a company or community, but within the users and those affected by technology)

19/🧵

A1 (@QuickResolute, contd.): fortunately, there are groups of us who are taking action and advocating for inclusion and representation within this field of tech

20/🧵

A1 (@vuisnotabot): we are just at the beginning of addressing marginalization wrt resources for visibility; systemic mismatches in identity and cultural values

21/🧵

A1 (@skoularidou): policymaker involvement and changemakers are necessary for marginalized groups to survive (in general and to have a sustainable identity in science and tech); (contd.)

22/🧵

A1 (@skoularidou, contd.): we need to sit down with policymakers in academia and industry so that they take serious account of our existence

Particular issues include fast-track hires and privileged hires (which run counter to fairness and inclusion)

23/🧵

A1 (@ShawnTsosie4): a key issue for us is protection of datasets (data sovereignty) and the revocability of data rights

Not all of these systems will be around forever and some of the resources may be hard to sustain and maintain

24/🧵

A1 (@jeffadoctor): a key question for indigenous peoples is acknowledgement of diversity, especially the sovereign distinction between different groups of us, ethnically and culturally; (contd.)

25/🧵

A1 (@jeffadoctor, contd.): a lot of techno-optimism is highly colonial and erases the logic of indigenous people's existence; tech is dual-use wrt potential decolonization and co-opting, and homogenization

26/🧵

A1 (@QuickResolute): this resonates with ethnocultural stereotyping, an issue that is salient to Latinx people from many countries and regions (contd.)

27/🧵

A1 (@QuickResolute, contd.): for instance, in AI, advertising to stereotyped and archetyped workers (e.g., agricultural workers) can lead to invalid assumptions

28/🧵

A1 (@skoularidou): also true of disabled persons wrt qualifications for promotion, or for example neurodivergent people and social interactions, and assumed capabilities and potential skills

29/🧵

Q2 (FRONTIERS IN EDUCATION): what do you see as the most interesting and meaningful facet of artificial intelligence education for coming years, coming out of recent technical developments and events (including the pandemic)?

30/🧵

A2 (@skoularidou): patients of the latter 21st century and beyond may regain more autonomy with respect to treatments (more choices and also more facilitative support from AI); e.g., drug design, assistive tech, procedures/surgeries

31/🧵

A2 (@vuisnotabot): improvements in AI (NLP, vision, automation) via awareness of diverse populations

32/🧵

A2 (@QuickResolute): despite controversy wrt big data nudging, health care is a big frontier in AI for good; in areas of education more broadly, people have tried to implement a lot of AI tools (contd.)

33/🧵

A2 (@QuickResolute, contd.): to better understand how an individual learns (within curricula and throughout their lifetime); another important facet of this is privacy (contd.)

34/🧵

A2 (@QuickResolute, contd.): finally, differentiated/individualized instruction through ML and other AI, this includes the present Covid situation and can perhaps help further education in the future

35/🧵

A2 (@ShawnTsosie4): it would be great to learn more about this because we are developing language resources for Kwak'wala and Makah and an AI code camp for Lakota youth

36/🧵

A2 (@jeffadoctor): things that we would like future AI developers and users, including indigenous youth: who is training AI? who is providing the data? (contd.)

37/🧵

A2 (@jeffadoctor, contd.): anthropological context; respect for sovereignty; awareness that indigenous people are alive and still being affected by the development and use of AI; centering the needs and rights of people whom data belongs to or comes from (contd.)

38/🧵

A2 (@jeffadoctor, contd.): educating people is different from training models; people can shift paradigms, while AI furthers a specific agenda that may belong to people who are out to profit or insensitive to the existence and potential of indigenous populations (contd.)

39/🧵

A2 (@jeffadoctor, contd.): preventing exploitation, colonization, and genocide

40/🧵

A2 (@QuickResolute): related example - using @duolingo to learn new languages; how data is sourced has a bearing

41/🧵

A2 (@jeffadoctor): indigenous data sovereignty issues - who represents communities and nations? what companies create "standards"? do the people represented agree? (dialects, etc.) (contd.)

42/🧵

A2 (@jeffadoctor, contd.): indigenous data has a collective nature that does not always admit individual representatives; this is a huge challenge and a paradox. e.g., I use @duolingo as an app and if not for these issues, would be very happy to work with it (contd.)

43/🧵

A2 (@jeffadoctor, contd.): a healthy relationship requires examination of these matters; many of our languages never had a text or were written down - so we’re translating a translation to a degree, into a computer system that may not recognize our characters

44/🧵

A2 (@vuisnotabot): Anglo-American centrism and Eurocentrism; even through these examples of texts that are not traditionally English or [a western European language], we can see biases and ramifications of educational standards

45/🧵

A2 (@jeffadoctor): Capitalism, colonialism, these need to be considered

46/🧵

A2 (@ShawnTsosie4): the potential for unfair data acquisition and data uses in AI is not the fault of any particular company, but more a reflection of state interests and actors (global governments)

47/🧵

Q3 (MENTORSHIP AND OUTREACH): what do see as your role in AI mentorship and outreach, and what support do you wish you had (i.e., what help are you seeking or would you most welcome in this dimension of DEI)?

48/🧵

A3 (@QuickResolute): creating and facilitating opportunities for Latinx researchers, including in LatAm; some such opportunities entail possible relocation to other countries and a brain drain (contd.)

49/🧵

A3 (@QuickResolute, contd.): ameliorating this effect includes seeking out mentors who may be non-Latinx allies, mid-career, or grad students mentoring undergrads (contd.)

50/🧵

A3 (@QuickResolute, contd.): our role has been a successful one to date; a current plan and need includes setting up tools to help mentors, and for building relationships between mentors and mentees, building technical and resource-based bridges (contd.)

51/🧵

A3 (@QuickResolute, contd.): and setting meaningful career progress goals

This includes recruiting pipelines between communities

52/🧵

A3 (@vuisnotabot): I would like to talk about 2-3 specific programs that @QueerinAI has, esp. for people who are applying to grad school.

(1) guidance for what information to include; mentorship for grad and scholarship/fellowship applicants (contd.)

53/🧵

A3 (@vuisnotabot, contd.):

(2) buddy system at conferences; experience of AI conference can be very overwhelming (kid in a candy store; many topics and ideas); building friendships and collegial connections (contd.)

54/🧵

A3 (@vuisnotabot, contd.): with respect to aid, what would help is increasing non-Western representation when it comes to mentors (contd.)

55/🧵

A3 (@vuisnotabot, contd.): plus outreach when it comes to mentorship participation and eligibility, especially for young researchers who are not at or near hotspots for research; inclusion of voices that are not specific to the global West/North is important

56/🧵

A3 (@jeffadoctor): we are trying to figure out how to mentor appropriately; we have our own jobs and needed; we want to make sure the process is not exploitative (contd.)

57/🧵

A3 (@jeffadoctor, contd.): we work with indigenous people, who have qualifications and knowledge that we seek; however, we acknowledge that they have their own learning process, needs, and traumas that we want to respect (contd.)

58/🧵

A3 (@jeffadoctor, contd.): We don't yet have the generational wealth (in many facets of resources) to provide everything at once; holding space for these is important.

Shoutout to the First Nations Tech Council @FN_TechCouncil: technologycouncil.ca

59/🧵

A3 (@ShawnTsosie4): I came into AI via a nontraditional route, as an algebraic number theorist rather than as a computer scientist; outreach starts with raising awareness and interest (contd.)

60/🧵

A3 (@ShawnTsosie4, contd.) a lot of indigenous youth don't participate in AI, not because they are not interested, but because they weren't told about it (contd.)

61/🧵

A3 (@ShawnTsosie4, contd.) this requires code camps as a start because it takes 10-15 years to build up to an AI workforce in research, education, industry (contd.)

62/🧵

A3 (@ShawnTsosie4, contd.) Not having to "leave home" is a big matter for indigenous people who want careers in this field (my path is similar to Michael's [@Obnoxious_Wolf's] in this regard in that we went through the non-indigenous education system for our background)

63/🧵

A3 (@jeffadoctor): another question for us is "what are ways that these pathways can be made conducive to our desired traditional lifestyles, without being exploited by labor seekers and/or the "savior complex?"

64/🧵

A3 (@vuisnotabot): see @outreachy - outreachy.org; they can be a good case study in how mentorship can increase the inclusion of underrepresented voices; pairing interns with mentors

65/🧵

A3 (@jeffadoctor): And @nativesintech - nativesintech.org

66/🧵

A3 (@vuisnotabot) @QueerinAI + oSTEM [@OUTinSTEM] financial aid & microgrants sites.google.com/view/queer-in-…

- end of panel -

67/🧵

Affinity group introduction by Angle Bush, founder of @BlackWomenInAI: blackwomeninai.com

"Educate, Engage, Embrace, Empower"

#AAAI2022 #AAAI #AffinityGroup #DiversityInAI

68/🧵

Third presentation: "Monitoring Diversity of AI Conferences: Lessons Learnt and Future Challenges in the DivinAI Project" - Isabelle Hupont (@ihupont), Emilia Gómez (@emiliagogu), Songül Tolan (@SonguelTolan), Lorenzo Porcaro (@porcaro_lorenzo), & Ana Freire (@ana__freire)

69/🧵