Embedding data ethics in the Institute for Geospatial Understanding Through an Integrative Discovery Environment (I-GUIDE)

Darch, Peter T.; Kong, Ivan; Abrams, Kyra M.

Permalink

https://hdl.handle.net/2142/119585 Copy

Description

Title

Embedding data ethics in the Institute for Geospatial Understanding Through an Integrative Discovery Environment (I-GUIDE)

Author(s)

• Darch, Peter T.
• Kong, Ivan
• Abrams, Kyra M.

Issue Date

2024

Keyword(s)

• data ethics
• AI ethics
• machine learning
• geospatial data
• sustainability science

Date of Ingest

2025-02-06T13:18:57-06:00

Abstract

Scholarly research using AI frequently involves reusing secondary data. A project can involve integrating multiple, heterogeneous datasets curated in diverse institutional, disciplinary, and cultural contexts according to divergent standards, policies, and norms. Such projects implicate multiple ethical issues, which are all-too-often side-lined. This poster presents our team’s work embedding data ethics in the Institute for Geospatial Understanding through an Integrative Discovery Environment (I-GUIDE), a five-year, $15-million project funded by the US National Science Foundation Harnessing the Data Revolution (HDR) program. I-GUIDE is developing an AI platform for sustainability researchers to analyse secondary geospatial datasets, with a goal of influencing decision-makers in government and industry. I-GUIDE is unique among HDR projects in funding a team of social scientists to address data ethics. This poster presents two of our projects: 1) understanding model bias to enhance transparency and accountability of AI-based analyses and decision-making; 2) studying how to intervene in work practices of I-GUIDE researchers to support ethical practice. These projects employ qualitative methods (observation, interviews, document analysis.) We have developed a theoretical framework informed by approaches in Library and Information Science, such as data curation lifecycles, documentation practices, and information-seeking behaviour. Using the I-GUIDE platform involves many ethical challenges. Data about people or human processes can raise privacy challenges and lead to discrimination if groups are under- or over-represented. Reuse of human subjects’ data poses challenges relating to consent. Using AI approaches raises concerns around bias, transparency, and accountability. Meanwhile, CARE principles emphasize issues relating to Indigenous data, such as sovereignty. Research using data not about people can also have ethical implications when used to inform policy- or decision-making impacting communities. Challenges can multiply due to divergent ethical perspectives as datasets are transferred across disciplinary, national, and cultural boundaries. In project #1, we are tracing the origins of the components of a major model employed by a particular group of researchers using the I-GUIDE platform. This research-in-progress is characterizing when, where, and how bias is introduced into this model. We combine qualitative interviews with researchers who developed components of this model with analysis of the scientific literature introducing these components. In project #2, we have found that researchers using the I-GUIDE platform are often concerned about ethical issues. However, they typically lack formal ethics training and struggle to identify how their practices impact ethical outcomes. Under pressure to meet deadlines and pursue career objectives, researchers frequently side-line ethical issues, viewing their responsibilities as purely technical while framing ethics as the concern of project management. Researchers want tools that are easy-to-integrate into their existing work practices. We have evaluated the suitability of 25 extant tools to support data ethics (e.g., checklists and data cards), adapting the most promising for IGUIDE researchers work practices. Trials of these tools are ongoing.

Publisher

Embedding Ethics in Geospatial Data Science, International Digital Curation Conference, Edinburgh, UK

Type of Resource

text

Genre of Resource

conference poster

Language

eng

Sponsor(s)/Grant Number(s)

National Science Foundation # 2118329

Owning Collections