Governance in ethical and trustworthy AI systems : Extension of the ECCOLA method for AI ethics governance using GARP
Agbese, Mamia; Alanen, Hanna-Kaisa; Antikainen, Jani; Halme, Erika; Isomäki, Hannakaisa; Jantunen, Marianna; Kemell, Kai-Kristian; Rousi, Rebekah; Vainio-Pekka, Heidi; Vakkuri, Ville (2023)
Agbese, Mamia
Alanen, Hanna-Kaisa
Antikainen, Jani
Halme, Erika
Isomäki, Hannakaisa
Jantunen, Marianna
Kemell, Kai-Kristian
Rousi, Rebekah
Vainio-Pekka, Heidi
Vakkuri, Ville
Wrocław University of Science and Technology
2023
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi-fe20230915126815
https://urn.fi/URN:NBN:fi-fe20230915126815
Kuvaus
vertaisarvioitu
© 2023 The Authors. Published by Wrocław University of Science and Technology Publishing House. This is an open access article under the CC BY license international.
© 2023 The Authors. Published by Wrocław University of Science and Technology Publishing House. This is an open access article under the CC BY license international.
Tiivistelmä
Background:
The continuous development of artificial intelligence (AI) and increasing rate of adoption by software startups calls for governance measures to be implemented at the design and development stages to help mitigate AI governance concerns. Most AI ethical design and development tools mainly rely on AI ethics principles as the primary governance and regulatory instrument for developing ethical AI that inform AI governance. However, AI ethics principles have been identified as insufficient for AI governance due to lack of information robustness, requiring the need for additional governance measures. Adaptive governance has been proposed to combine established governance practices with AI ethics principles for improved information and subsequent AI governance. Our study explores adaptive governance as a means to improve information robustness of AI ethical design and development tools. We combine information governance practices with AI ethics principles using ECCOLA, a tool for ethical AI software development at the early developmental stages.
Aim:
How can ECCOLA improve its robustness by adapting it with GARP® IG practices?
Methods:
We use ECCOLA as a case study and critically analyze its AI ethics principles with information governance practices of the Generally Accepted Recordkeeping principles (GARP®).
Results:
We found that ECCOLA’s robustness can be improved by adapting it with Information governance practices of retention and disposal.
Conclusions:
We propose an extension of ECCOLA by a new governance theme and card, # 21.
The continuous development of artificial intelligence (AI) and increasing rate of adoption by software startups calls for governance measures to be implemented at the design and development stages to help mitigate AI governance concerns. Most AI ethical design and development tools mainly rely on AI ethics principles as the primary governance and regulatory instrument for developing ethical AI that inform AI governance. However, AI ethics principles have been identified as insufficient for AI governance due to lack of information robustness, requiring the need for additional governance measures. Adaptive governance has been proposed to combine established governance practices with AI ethics principles for improved information and subsequent AI governance. Our study explores adaptive governance as a means to improve information robustness of AI ethical design and development tools. We combine information governance practices with AI ethics principles using ECCOLA, a tool for ethical AI software development at the early developmental stages.
Aim:
How can ECCOLA improve its robustness by adapting it with GARP® IG practices?
Methods:
We use ECCOLA as a case study and critically analyze its AI ethics principles with information governance practices of the Generally Accepted Recordkeeping principles (GARP®).
Results:
We found that ECCOLA’s robustness can be improved by adapting it with Information governance practices of retention and disposal.
Conclusions:
We propose an extension of ECCOLA by a new governance theme and card, # 21.
Kokoelmat
- Artikkelit [2902]