Page 263 - Kaleidoscope Academic Conference Proceedings 2024
P. 263

ADVANCING TRUSTWORTHY AI FOR SUSTAINABLE DEVELOPMENT:
                   RECOMMENDATIONS FOR STANDARDISING AI INCIDENT REPORTING

                                                           1
                                             Avinash, Agarwal and Manisha, Nene 2
                1
                 (avinash.70@gov.in) Telecommunication Engineering Centre, Ministry of Communications, New Delhi, India
                     2
                      (mjnene@diat.ac.in) Defence Institute of Advanced Technology, Ministry of Defence, Pune, India



                              ABSTRACT                        Several studies highlight that if not deployed responsibly and
                                                              ethically, AI could impede the achievement of the UN SDGs
           The increasing use of AI technologies has led to increasing  [5, 6]. For instance, algorithmic biases in hiring processes
           AI incidents, posing risks and causing harm to individuals,  could exacerbate inequalities in employment opportunities,
           organizations, and society.  This study recognizes and  thereby impeding progress toward SDG 8 (Decent Work and
           addresses the lack of standardized protocols for reliably  Economic Growth) and SDG 10 (Reduced Inequalities) [7].
           and comprehensively gathering such incident data crucial
           for preventing future incidents and developing mitigating  Principles of Responsible AI, such as those proposed by
           strategies.  Specifically, this study analyses existing  Organization for Economic Co-operation and Development
           open-access AI-incident databases through a systematic  (OECD), emphasize inclusive growth, sustainability, fairness,
           methodology and identifies nine gaps in current AI incident  transparency, robustness, and accountability of AI systems
           reporting practices. Further, it proposes nine actionable  [8]. Compliance with these principles can ensure that AI
           recommendations to enhance standardization efforts to  systems aid and do not hamper the achievement of UN
           address these gaps. Ensuring the trustworthiness of enabling  SDGs. Standards, benchmarks, and standardized assessment
           technologies such as AI is necessary for sustainable digital  procedures are needed to ensure that AI systems meet
           transformation.  Our research promotes the development  the responsible AI principles [9, 10].  Comprehensive
           of standards to prevent future AI incidents and promote  data collected through different AI lifecycle stages and
           trustworthy AI, thus facilitating achieving the UN sustainable  deployments in diverse scenarios drives the assessment of
           development goals.  Through international cooperation,  compliance with these responsible AI principles.
           stakeholders can unlock the transformative potential of AI,
           enabling a sustainable and inclusive future for all.  Learning from past AI incidents is a crucial way to avoid
                                                              repeat incidents. The aviation industry has well-established
             Keywords - AI incident database, AI harm, adversarial  protocols for collecting aviation incident-related data.
                  attack, SDG, standardization, sustainability  Systematically collecting and analyzing details of aviation
                                                              incidents have resulted in continuous product improvement
                         1.  INTRODUCTION                     and mitigation strategies, leading to a drastic reduction in
                                                              aviation accidents [11, 12]. Similarly, cybersecurity incident
           The proliferation of AI technologies across diverse domains  reporting is well-established and supported by regulations in
           has led to rapidly increasing AI incidents ranging from  many countries [13].
           algorithmic biases and deepfakes to system failures and
           unintended consequences.  These incidents pose risks to  Transparent  disclosure  of  incidents,  comprehensive
           individuals, organizations, and society, thus undermining  compilation of AI incident data, and their systematic analysis
           overall trust and confidence in AI technologies. Recognizing  can provide crucial data for developing mitigation strategies
           the importance of addressing these risks, stakeholders  and promoting the deployment of trustworthy AI [14].
           are increasingly focusing on identifying, analyzing, and  Against the backdrop of the UN SDGs, which seek to
           mitigating AI-related risks and harms.             address pressing global challenges and promote sustainable
                                                              development, the need for standardized AI incident reporting
           Sustainable digital transformation, driven by innovative  becomes even more pronounced. This paper identifies and
           technologies such as AI, can accelerate progress towards  addresses the critical gap in the availability of standards
           the United Nations’ Sustainable Development Goals (SDGs),  and protocols for systematic AI incident reporting and data
           for example by enhancing access to quality healthcare (SDG  sharing.
           3) [1], education (SDG 4) [2], managing water crisis and
           sanitation (SDG 6) [3], and climate change adaptation  In light of these considerations, this paper explores the
           (SDG 13) [4], among other benefits. However, realizing  intersection of AI incident reporting, sustainable digital
           this potential requires a concerted effort to ensure that AI  transformation, and the UN SDGs. By analyzing the current
           technologies are deployed responsibly and ethically, with due  state ofAI incidentreporting, identifying gaps andchallenges,
           consideration for their societal and environmental impacts.  and proposing recommendations for improvement, this




            978-92-61-39091-4/CFP2268P @ITU 2024          – 219 –                                   Kaleidoscope
   258   259   260   261   262   263   264   265   266   267   268