1 of 25

A data ethics and data justice approach for AI-Enabled OER

​

Dr Javiera Atenas Leo Havemann

University of Suffolk UCL & OU

#OER24

DOI 10.5281/zenodo.10891946

2 of 25

Context

This exploratory research focuses on the need to equip educators with a critical understanding of ethical issues in the AI space such as algorithmic discrimination so they can anticipate and respond to issues related to the collection, processing and use of AI in the development of OER.

3 of 25

Open Education and AI

  • OE advocates for reducing the barriers to access and participation, widening learning opportunities while democratising education.
  • This involves OEP which promotes collaboration and sharing good, effective, creative and innovative practices, and the use and creation of OER, which are currently defined as “teaching and learning materials that are freely available to use, adapt, and share”.

4 of 25

Open Education and AI

  • This definition does not address the possibilities of AI-enabled OER.

​

  • AI services now present opportunities to create, adapt, personalise and contextualise resources in all shapes and forms.

​

  • It’s even been suggested that OER could consist just of prompts - AI can generate the rest.

​

  • We must considering that the risks implied in this process, due to the biases encrusted into data- and algorithm- driven systems.

5 of 25

Ethical considerations

  • The development of AI-enabled OER must address issues related to bias and discrimination: the datasets used to inform generative AI platforms (‘training data’) can amplify social inequalities and reinforce discriminatory practices, leading to biased and unfair treatment or portrayal of marginalised communities.

​

  • Thus, AI-enabled OER need to be designed to support learners considering elements of data justice and data ethics to ensure AI-enabled OER are inclusive, representative and challenge power inequalities instead of amplifying them.

6 of 25

AI and Machine Learning

Machines are trained by humans, and if biased information or data reflecting existing inequities is fed into machine learning programs, the algorithms will learn and perpetuate those biases, leading to discrimination.

​

We also need to consider that the much of the human work of training generative AI systems is powered by underpaid workers globally, performing repetitive tasks under precarious labour conditions, often recruited from impoverished populations.

7 of 25

Bias in AI

Fuente: Datacamp Data Literacy Richie Cotton

Data Demystified: The Different Types of AI Bias https://www.datacamp.com/blog/data-demystified-the-different-types-of-ai-bias

8 of 25

Bias in AI – an historical example

9 of 25

Bias in AI

Fuente: Mary Reagan https://towardsdatascience.com/understanding-bias-and-fairness-in-ai-systems-6f7fbfe267f3

10 of 25

A very simple example

11 of 25

A very simple example

12 of 25

A very simple example

13 of 25

Ethical issues of AI

proposed categorization of ethical issues in the field of AI by Huang et al. (2023).

14 of 25

One approach: Data feminism as a framework for AI enabled OER development

  • Examine power. Data feminism begins by analysing how power operates in the world. 
  • Challenge power. Data feminism commits to challenging unequal power structures and working toward justice. 
  • Elevate emotion and embodiment. Data feminism teaches us to value multiple forms of knowledge, including the knowledge that comes from people as living, feeling bodies in the world. 
  • Rethink binaries and hierarchies. Data feminism requires us to challenge the gender binary, along with other systems of counting and classification that perpetuate oppression. 
  • Embrace pluralism. Data feminism insists that the most complete knowledge comes from synthesising multiple perspectives, with priority given to local, Indigenous, and experiential ways of knowing. 
  • Consider context. Data feminism asserts that data is not neutral or objective. It is the product of unequal social relations, and this context is essential for conducting accurate, ethical analysis. 
  • Make labour visible. The work of data science, like all work in the world, is the work of many hands. Data feminism makes this labour visible so that it can be recognised and valued. 

15 of 25

DATA FEMINISM PRINCIPLES FOR AI-ENABLED OER DEVELOPMENT

Examine power

Who are the dominant voices informing your OER? (e.g. theorists, leaders)

Where do your good practices come from?

Challenge power

Are certain theorists, scholars, practitioners or groups not mentioned in your OER?

How are vulnerable or diverse groups portrayed in your OER?

Elevate emotion and embodiment

Does your OER refers to lived experiences?

Does your OER fairly reflect and portray people/groups?

Rethink binaries and hierarchies

Does your OER reflects diversity?

Are you using inclusive language in your design?

Embrace pluralism

Are other knowledges represented in your OER?

Do you have a bias mitigation strategy for the development of your OER?

Consider context

Does your OER address unequal social relations

Does your OER address data and research ethics principles

Make labour visible

Does your OER acknowledges issues related with copyright breaches and misuse of works

Does your OER addresses issues of labour rights and conditions in terms of who is training the AI?

Javiera Atenas, 2024

16 of 25

GUIDELINES TO CONSIDER

The European Commission's Trustworthy AI Guidelines

Human agency and oversight

Diversity, non-discrimination, and fairness

Societal and environmental well-being

UNESCO Recommendation on the Ethics of AI

​

Inclusiveness

Accountability

Fairness and non-discrimination

Privacy and data protection

The IEEE Global Initiative on Ethics of Autonomous and Intelligent Systems

Ensuring the well-being of humanity

Transparency

Accountability

The Asilomar AI Principles

Broadly distributed benefits

Long-term safety

The Montreal Declaration for a Responsible Development of Artificial Intelligence

Ensure the well-being of all stakeholders

Avoid harm and minimise risks

Ensure fairness and justice in AI systems

Data Justice

Make visible community-driven needs, challenges, and strengths

be representative of community

treat data in ways that promote community self-determination

Javiera Atenas, 2024

17 of 25

A data ethics approach for openness – Atenas, Havemann and Timmermann, 2023

18 of 25

How to ethically develop AI enabled OER?

Here are some questions we may like to answer before developing an AI enabled OER

  • Do I really need to use AI for this?
  • What is outside of my control and what can I do about that?
  • Am I familiar with the platform(s) I will be using?
  • Have I thought about how others may be portrayed in my OER?
  • Are there any other open resources I could be manually remixing to mitigate the environmental impact?

19 of 25

How to ethically develop AI enabled OER?

Here are some questions we may like to answer before developing an AI enabled OER

  • Am I enabling critical thinking and critical literacies through my OER?
  • Am I acknowledging and attributing others in my OER considering that the information I have collected does not come from spontaneous generation?

20 of 25

Rationalising our decision-making processes on AI enabled OER development

21 of 25

Learning Resources Design Guidance – Prototype V1

Dr Javiera Atenas

Professor Nicholas Caldwell

University of Suffolk

22 of 25

How to ethically develop AI enabled OER?

This decision making tool aims at supporting educators to make effective and ethical decisions in the development of learning resources (copyrighted or open) ensuring they understand the challenges of legally using 3rd party content and also, ethically and effectively use AI to generate resources that are inclusive, contextual, accurate and reflect the diversity of knowledge.

​

You can access the Learning Resources Design Guidance tool here

​

UOS LO-OER decision tool

​

23 of 25

Advancing our research

Our aim is to understand, in praxis and in the light of the review of academic literature, which are the key competencies needed to scaffold the complex formative design and implementation of teaching, learning and assessment processes in our datafied society, with specific reference to the field of Higher Education.

​

You can access our survey here

​

https://forms.office.com/e/2x4j1KenBy

​

24 of 25

Making labour visible.

For this presentation I like to acknowledge the work and ideas of fellow OER folks and critical educators

  • Anne-Marie Scott
  • Leo Havemann
  • Chrissi Nerantzi
  • Lorna Campbell
  • Frances Bell
  • Davor Orlic
  • Wayne Holmes

​

​

  • Priscila Gonzales
  • Ben Williamson
  • Catherine Cronin
  • Rob Farrow
  • Daniel Villar

​

And also, we have sought inspiration in the work of

Catherine D’Ignazio

The data justice lab

​

25 of 25

Some resources you may find interesting

  • Atenas, J., Havemann, L., & Timmermann, C. (2023). Reframing data ethics in research methods education: A pathway to critical data literacy. International Journal of Educational Technology in Higher Education, 20(1), 11. https://doi.org/10.1186/s41239-023-00380-y
  • Ball, S. J. (2015). Education, governance and the tyranny of numbers. Journal of Education Policy, 30(3), 299–301. https://doi.org/10.1080/02680939.2015.1013271
  • Barocas, S., & Selbst, A. D. (2018). Big Data’s Disparate Impact. SSRN Electronic Journal, 671, 671–732. https://doi.org/10.2139/ssrn.2477899
  • Es, K. V., & Schäfer, M. T. (Eds.). (2017). The Datafied Society: Studying Culture through Data. Amsterdam University Press. http://www.oapen.org/search?identifier=624771

​

​