An Ecosystem Approach to Ethical AI and Data Use: Experimental Reflections

“language as well as a normative regulatory frame requiring shared responsibility resonates with Habermas’s discourse ethics paradigm [18].”

“In response to rising concerns of unethical AI development, the prevailing approach from governments, intergovernmental organisations, and big tech firms has been to roll out a ‘new’ vocabulary of ethics and responsibility [2].”

“Private companies like Google, Microsoft, IBM and Tencent have taken the lead in publishing their own ethics documents and principles [2], [5].”

“level and abstract, making it difficult for practitioners and technicians to see the their relevance in their daily activities [4], [9].”

“continue to be grappled with even as the field is shifting from the ’what of AI ethics’ to a more operational ’how’ [8].”

“Work incorporating AI developers’ own perspectives in the ethics debate remains currently limited and relatively under explored [9].”

“Veale, Van Kleek and Binns interviewed public sector machine learning practitioners working across five countries to understand how they were putting considerations of fairness and accountability into their everyday practices [11].”

“The concept of AI for Social Good (AI4SG) has been adopted widely by developers within the AI community [1];”

“There are also examples of this discourse of negotiation and meaning sharing in the preparation of best practice guidelines when codes of conduct are struck to encourage internal industry regulatory compliance [19].”

“and approaches the discussion from an understanding of what they need; second, like Orr and Davis’s work [12], it helpfully situates the discussion of ethics and responsibility beyond”

“organisationally supported with adequate tools or methods for addressing them in their work [10].”

“Dialogue, especially moderated by outsiders, becomes a form of provocation that interrupts the everyday practices of AI developers and helps to question practices which may have been otherwise taken for granted [21].”

“1 3 is in constant danger of being hijacked by particular interests, most notably the interests of large corporations who have a vested interest in using ethical rhetoric to avoid regulation and deflect scrutiny (Nemitz 2018; Coeckelbergh 2019; Mittelstadt 2019; Findlay and Seah 2020).”

“The metaphor of ecosystems to describe AI and its broader social and ethical consequences has already been employed widely by scholars [53] as well as policymakers.”

“Moreover, conceptualizing the networked interactions of AI actors as ecosystems is well established in the academic literature (Findlay & Seah, 2020; Stahl, 2021, p. 81) and policy and strategy papers (e.g., European Commission, 2020; OECD, 2019).”

“The principle of Findability is important for ensuring that data and resources are not only discoverable but also readily accessible [18].”

“AI teams repeatedly report that their organizations ’ fi nancial requirements, expectations, and compensation structures are not compatible with the investment necessary to fully grapple with the ethical problems their AI products pose (Findlay and Seah, 2020; Rakova et al., 2021).”

“Consequently, there are frequent references to “AI ecosys-tems”, particularly in interventions that highlight the ethical complexities of AI (see, e.g., Findlay and Seah 2020; Fjeld et al. 2020; UK Government 2021; UNESCO 2020).”

“For instance, researchers found that AI practitioners need immense guidance to implement RAI practices related to dataset documentation, harm identification, and explainability techniques [29, 30, 42, 54, 75, 78, 96, 108, 109].”

“Drawing inspiration from a range of interdisciplinary research [Holländer et al. 2019; Findlay and Seah 2020; Hoff and Bashir 2015; Kraus 2020], this paper proposes a community-focused approach to build trust between users and key decision-makers in the wider robotics ecosystem.”