1 of 30

Data Ethics

Katerina Allmendinger

Nickoal Eichmann-Kalwara

Data Camp 2024

Image credit: iterations of “a fluffy black cat sleeping on top of datasets, watercolor” - Unknown, and Nickoal+DALL-E, 2023.

For us, talking about data ethics is grounded in indigenous knowledge frameworks and critical data studies, towards process of working to dismantle the ongoing legacies of settler colonialism, towards harm reduction, and working towards a free world with collective autonomy.

With this and care in mind, we acknowledge we are witnessing the violence settler colonialism in real time, currently there are multiple genocides of indigenous peoples happening. And, i want to remind you, Tune into your mind and body for how you tune in and engage today. You’re welcome to just listen, share reflections as we go in the chat, whatever is best for you. Thank you for your presence and participation with us.

2 of 30

Data are constructed - they are the products of biased human choices and perspectives.

Data Camp 2024

3 of 30

Data Ethics and the Research Lifecycle

What does ethics [data / research] mean to you?

At which stages should you critically engage with your data?

Data Camp 2024

4 of 30

Data Ethics and the Research Lifecycle

Uphold applicable statutes, regulations, professional practices, and ethical standards
Respect the public, individuals, and communities
Respect privacy and confidentiality
Act with honesty, integrity, and humility
Hold oneself and others accountable
Promote transparency

Based on Federal Data Strategy

Data Camp 2024

5 of 30

Data as Texts/Artifacts

Data are not neutral or objective.
Data do not represent a self-evident, discoverable Truth.

Rather,

Data are constructed, the products of biased human choices and perspectives.
The interpretation, use, and analysis of data, like any text, is shaped by our biases and perspectives.

Data Camp 2024

6 of 30

Code4Lib Journal, December 2023

7 of 30

8 of 30

https://www.go-fair.org/fair-principles/

https://www.gida-global.org/care

Data Camp 2024

machine actionability

social actionability

there is a growing push for sharing and reusing data by humans and machines. The FAIR principles are increasingly being used as a framework for this. You’ll learn more about this in later sessions, but the summarize,

However appropriate sharing and reuse does require awareness of and application of governance and stewardship mechanisms that are relevant for indigenous data.

Social actionability as social and collective care

The CARE Principles for Indigenous Data Governance were drafted at the International Data Week and Research Data Alliance Plenary co-hosted event “Indigenous Data Sovereignty Principles for the Governance of Indigenous Data Workshop,” 8 November 2018, Gaborone, Botswana.

Global Indigenous Data Alliance (GIDA)

(F)indable and (A)ccessible and (I)nteroperable and (R)e-usable
(C)ollective benefit: Data ecosystems should be designed and function in ways that enable indigenous peoples to derive benefit from the data.
(A)uthority to control: Indigenous Peoples’ rights and interests in Indigenous data must be recognised and their authority to control such data be empowered.
(R)esponsibility: Those working with Indigenous data have a responsibility to share how data is used to support Indigenous Peoples’ self-determination and collective benefit.
(E)thics: Indigenous Peoples’ rights and well-being should be the primary concern at all stages of the data life cycle and across the data ecosystem.

talking about data ethics = grounded in indigenous knowledge frameworks

9 of 30

https://www.gida-global.org/care

Data Camp 2024

10 of 30

CARE Data Principles

From the perspective and ethical values of the involved community, what are the potential benefits and harms of collecting, using, or sharing these data now and in the future?
Does the data play into harmful stereotypes and stigmas, or does it empower the community to assert their rights and interests?
Are the provenances, purposes, and use limitations stated in the metadata or other documentation?
Do metadata standards incorporate the (Indigenous) communities values and concepts?

Data Camp 2024

11 of 30

Consent

Privacy

Care work

Harm Reduction

Sovereignty

Data Camp 2024

12 of 30

Power

Privacy

Trust

Justice

Data Camp 2024

13 of 30

Power

Who is collecting and analyzing the data?
What’s counted, or not?
How and when something is counted and why?
Current social and political concerns as influencers?
Right to be Forgotten (GDPR)

Data Camp 2024

14 of 30

Privacy & Security

Consent: Permission to use these data (ownership?)
Anonymization: removing sensitive and identifiable information
Confidentiality: Whose/what information is protected and for how long?

Data Camp 2024

15 of 30

Trust

Forms of Knowledges
Accountability
Documentation
Reproducibility

Data Camp 2024

16 of 30

Justice

Social impact
Missing and counter data
Protection
Co-liberation

Data Camp 2024

17 of 30

Ethics for Artificial Intelligence (AI)

Data Camp 2024

18 of 30

So while AI is novel, it often seems to continue long-standing paradigms of technology in the service of capital … scholarship suggests that the human harms documented in recent AI-driven initiatives are not merely “teething problems,” but part of a broader paradigm of capitalist and colonialist values at the core of our current economic and technological systems. (Munn, 2023)

Data Camp 2024

https://doi.org/10.1007/s00146-023-01636-x

One other frame I want to use before we dive in is to look at AI ethics holistically and realistically. Because these are tools created by imperfect humans within systems of inequality and oppression, the creators’ cultural contexts are going to be encoded, and then can be exacerbated by the wide use and re-use of AI-generated content. It’s important to be intentional with our language - if we’re attributing sentience or bias to AI itself we’re putting ourselves at a disadvantage and further disempowering ourselves in addressing ethical issues in AI. So we really need to consider the broader societal context when we’re looking at AI tools.

Because these are a continuation of data ethics with concerns around:

Bias and misrepresentation
Exploitation of people and environment
Creator and end user privacy and confidentiality
Related to all of these things but of special concern - its use to generate and spread misinformation
AI is especially vulnerable to the butterfly effect, of small changes being introduced to the system and then being amplified, due to the nature of LLM based generative AI tools

19 of 30

a biological scientist

a chief executive officer

Image: Quote excerpt from Gross, N. (2023)

Bias

Image from Sun, L. et al. (2023)

20 of 30

Exploitation

"Dragon Cage" by Greg Rutkowski.

Data Camp 2024

AI-generated art by Sandy60

Two examples here to demonstrate how genAI can exploit culture and livelihood. Also consider environmental implications of the huge data centers behind AI tools. This energy consumption is in large part due to training the LLMs, but answering queries also uses energy. AI already uses as much energy as a small country. It’s only the beginning

Indigenous knowledges informing 'machine learning' could prevent stolen art and other culturally unsafe AI practices (theconversation.com)

"Aboriginal art" by Sandy60 - This is art created in Stable Diffusion 1.5 with the text prompt "Aboriginal art"
So this image is completely stripped of context - we could assume or guess that the person who generated this image is not Aboriginal, but we don’t know. We don’t even know if this was created with an intention to represent a specific art style The way that these images are shared within the social

Artists File Class Action Lawsuit Against AI Image Generator Giants -

21 of 30

How is user input data treated? Is it stored? Anonymized? Shared with other companies?
What are the safeguards against malicious attacks?
Are these practices transparent and shared with users for informed consent?

Privacy and Security

Image credit: Photo by Lianhao Qu on Unsplash

Image credits: Photo by Lianhao Qu on Unsplash

A very quick few words on what is a huge topic - privacy and security, and by extension, also copyright. This impacts us as both end users of GenAI tools and as creators of information available online

First, privacy and security: hopefully, if you’re using AI tools for something like coding sensitive research data, you already know the tool’s data security policies before you go through your IRB process. It’s something to keep in mind - how will the privacy and data security policies of a tool fit into your research process?

Microsoft Copilot as an example - there is the freely available version and there’s the enterprise version.

Here’s a can of worms - copyright and generative AI. This is getting beyond my scope, so I’m opening the can of worms only enough to acknowledge that these privacy and security concerns can impact our careers and livelihoods. Keeping the lid safely on that can of worms by noting that 1) I’m not a lawyer and 2) much of this is yet to be decided in the courts.

Some reading linked in the slides that go into this further, but I also recommend working with colleagues - if you have a law school, library, campus legal counsel, those are all people who you might be able to work with to learn more.

Copyrights, Professional Perspective - IP Issues With AI Code Generators from Bloomberg News

Generative Artificial Intelligence and Copyright Law from the Congressional Research Service

22 of 30

Misinformation

From left: 1) AI-generated image that falsely depicts Kamala Harris and Donald Trump smiling and high-fiving 2) Social media post from Donald Trump sharing an AI-generated image that

AI-Generated Image

23 of 30

Using Generative Artificial Intelligence

Questions to consider:

How was this tool created and how is it maintained? By whom?
Who can use this tool and for what purposes?
How does my use of this tool impact me, my immediate communities, creators of input data, larger social structures, ecosystems, and intersections of those groups?
… and what do we do if we cannot answer these questions?

Data Camp 2024

As we transition into talking about AI ethics, I want to share some framing questions. These concerns will look familiar because AI uses and produces data.

How was this tool created? How is it maintained? By whom?

Scope and treatment of the corpus - where an opaque or so-called “black box” tool can be especially problematic, as we can’t always determine this

What types of data are included? How recent?
How has the corpus been trained to exclude culturally sensitive information, personally identifiable information where someone is at risk of harm?
Are changes made when they are shown to be necessary?

Cultural context of creators

Has this AI tool been designed or adapted to fit within a specific cultural context?
Because Silicon Valley is not a blueprint that fits everyone everywhere – AI must be adapted to the codes and practices of cultures when data sovereignty applies

Stated purpose and values - Of the responsible company/creators
Also important to consider this, because if use of this tool causes harm, we need to be able to advocate for change and for addressing the harm caused.

How does my use of this tool impact people and land?

Data sovereignty and safety - is it culturally appropriate for people outside a culture to have this data? Has personally identifiable information been scrubbed, where necessary for safety?
Privacy and confidentiality - How is your use of the tool and your inputs saved?
Responsible scholarship - Attribution of work, ability to
Environmental costs - to train, maintain, and use the tool

And what do we do if we cannot answer these questions? We’ll talk about that later.

We won’t always have the same questions to ask, because our contexts are different and may change. My main goal in voicing these is to keep our minds open and curious, and to consider our role in the content output from AI tools. My vantage point here is very much as a learner, and these questions are directly inspired by the work of Indigenous scholars around AI ethics and broader technology ethics. If we’re working with students, I think it’s important to bring up these questions to help them choose a tool to understand whether it will work for them.

24 of 30

How we work towards Ethical AI

Frameworks must center Indigenous knowledges and experiences:

Indigenous Protocol and Artificial Intelligence Position Paper (Lewis et al., 2020)
The Five Tests of Māori scholar Sir Hirini Moko Mead (Munn, 2023)

Be a continual learner - have curiosity towards information you encounter (and generate)
Is there ethical AI under capitalism?

Data Camp 2024

I’m trying to inject a little humor here, because this is a lot.

My last thought here is really coming from the work of Hirini Moko Mead, and I recommend reading the article linked here where Luke Munn considers how Mead’s tests apply to AI. Mead talks about applying precedent - he asks “Is there some event in our traditions that might help us understand the issue and help frame a response to it?” - and we do have the precedent, we know that we’ve been grappling with these ethical questions for many years before these technologies were developed. So we

So I encourage all of us to be continual learners and build community - because we’re seeing development of AI from new companies and continual updates to existing AI, it can be a lot to keep up with, and we can’t do it alone. Community work will be where we see the most progress. So I’m recommending curiosity and a willingness to learn alongside others, which is exactly what we’re hoping this session has been today. These forms of learning are of course dependent on your time and capacity and how you learn - but some possibilities here are to read investigative journalism, talk with your colleagues, read privacy policies and technical papers, read secondary literature that delves into AI ethics. There’s so much out there. Thank you - I’ll turn it back to Nickoal.

25 of 30

Transformative and radical uses of AI

A note of hope …

Data Camp 2024

26 of 30

Data are constructed - they are the products of biased human choices and perspectives.

Data Camp 2024

27 of 30

Reflections

What are examples of data ethical concerns that you encounter in your daily work, scholarship, and/or teaching?
In what ways do you think power is reflected in your data?
What data are you missing? Was it intentional? How might this missing data be beneficial?
How might thinking about your data as a text/artifact change how you relate to, document, and share your work?

28 of 30

References

Carroll, SC, Garba, I, Figueroa-Rodríguez, OL, Holbrook, J, Lovett, R, Materechera, S, Parsons, M, Raseroka, K, Rodriguez-Lonebear, D, Rowe, R, Sara, R, Walker, JD, Anderson, J and Hudson, M. The CARE Principles for Indigenous Data Governance. Data Science Journal (2020). https://doi.org/10.5334/dsj-2020-043
Wilkinson, M., Dumontier, M., Aalbersberg, I. et al. The FAIR Guiding Principles for scientific data management and stewardship. Sci Data 3, 160018 (2016). https://doi.org/10.1038/sdata.2016.18
Munn, L. The five tests: designing and evaluating AI according to indigenous Māori principles. AI & Society (2023). https://doi.org/10.1007/s00146-023-01636-x
Iliadis, A., & Russo, F. (2016). Critical data studies: An introduction. Big Data & Society, 3(2). https://doi.org/10.1177/2053951716674238
D'Ignazio, C. & Klein, L. Data Feminism. 2020. https://data-feminism.mitpress.mit.edu/
Fair Principles. GO FAIR. https://www.go-fair.org/fair-principles/
Open Letter to the Code4Lib Journal Editorial Board, December 2023. https://docs.google.com/document/d/1c-ADFaum9pB-W4AWQIxE6BPyVJs8goMlrVLoC00ucKc/edit?usp=sharing

29 of 30

References

Center for Countering Digital Hate. Fake Image Factories: How AI image generators threaten election integrity and democracy. https://counterhate.com/research/fake-image-factories/
Sun, L., Wei, M., Sun, Y., Suh, Y. J., Shen, L., & Yang, S. (2023). Smiling Women Pitching Down: Auditing Representational and Presentational Gender Biases in Image Generative AI (arXiv:2305.10566). arXiv. http://arxiv.org/abs/2305.10566
Gross, N. (2023). What ChatGPT Tells Us about Gender: A Cautionary Tale about Performativity and Gender Biases in AI. Social Sciences, 12(8), Article 8. https://doi.org/10.3390/socsci12080435
Mylrea, M., & Robinson, N. (2023). Artificial Intelligence (AI) Trust Framework and Maturity Model: Applying an Entropy Lens to Improve Security, Privacy, and Ethical AI. Entropy, 25(10), 1429. https://doi.org/10.3390/e25101429
Ferrara, E. (2024). The Butterfly Effect in artificial intelligence systems: Implications for AI bias and fairness. Machine Learning with Applications, 15, 100525. https://doi.org/10.1016/j.mlwa.2024.100525
Meighan, P. J. (2021). Decolonizing the digital landscape: The role of technology in Indigenous language revitalization. AlterNative: An International Journal of Indigenous Peoples, 17(3), 397–405. https://doi.org/10.1177/11771801211037672

30 of 30

Data Ethics

Katerina Allmendinger

katerina.allmendinger@colorado.edu

Nickoal Eichmann-Kalwara

nickoal.eichmann@colorado.edu

Data Camp 2024

THANKS!