Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
41 tokens/sec
GPT-4o
60 tokens/sec
Gemini 2.5 Pro Pro
44 tokens/sec
o3 Pro
8 tokens/sec
GPT-4.1 Pro
50 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Simulating Policy Impacts: Developing a Generative Scenario Writing Method to Evaluate the Perceived Effects of Regulation (2405.09679v2)

Published 15 May 2024 in cs.CL and cs.AI

Abstract: The rapid advancement of AI technologies yields numerous future impacts on individuals and society. Policymakers are tasked to react quickly and establish policies that mitigate those impacts. However, anticipating the effectiveness of policies is a difficult task, as some impacts might only be observable in the future and respective policies might not be applicable to the future development of AI. In this work we develop a method for using LLMs to evaluate the efficacy of a given piece of policy at mitigating specified negative impacts. We do so by using GPT-4 to generate scenarios both pre- and post-introduction of policy and translating these vivid stories into metrics based on human perceptions of impacts. We leverage an already established taxonomy of impacts of generative AI in the media environment to generate a set of scenario pairs both mitigated and non-mitigated by the transparency policy in Article 50 of the EU AI Act. We then run a user study (n=234) to evaluate these scenarios across four risk-assessment dimensions: severity, plausibility, magnitude, and specificity to vulnerable populations. We find that this transparency legislation is perceived to be effective at mitigating harms in areas such as labor and well-being, but largely ineffective in areas such as social cohesion and security. Through this case study we demonstrate the efficacy of our method as a tool to iterate on the effectiveness of policy for mitigating various negative impacts. We expect this method to be useful to researchers or other stakeholders who want to brainstorm the potential utility of different pieces of policy or other mitigation strategies.

PDF HTML Abstract
Ai Sparkles Streamline Icon: https://streamlinehq.com Summarize File Document Download Save Streamline Icon: https://streamlinehq.com PDF File Document Download Save Streamline Icon: https://streamlinehq.com Markdown Bookmark Chat Bubble Oval Streamline Icon: https://streamlinehq.com Chat (Pro)
Definition Search Book Streamline Icon: https://streamlinehq.com
References (60)
  1. Gpt-4 technical report. arXiv preprint arXiv:2303.08774.
  2. Generative AI in education and research: Opportunities, concerns, and solutions. Journal of Chemical Education, 100(8): 2965–2971.
  3. A review of scenario planning. Futures : the journal of policy, planning and futures studies, 46: 23–40. ISBN: 0016-3287.
  4. The anticipatory paradigm. AI Magazine, 44(2): 133–143.
  5. Anthropic. 2024. Claude 3 haiku: our fastest model yet.
  6. The Moral Machine experiment. Nature, 563(7729): 59–64.
  7. Barnett, J. 2023. The ethical implications of generative audio models: A systematic literature review. In Proceedings of the 2023 AAAI/ACM Conference on AI, Ethics, and Society, 146–161.
  8. Crowdsourcing impacts: exploring the utility of crowds for anticipating societal impacts of algorithmic decision making. In Proceedings of the 2022 AAAI/ACM Conference on AI, Ethics, and Society, 56–67.
  9. Benjamin, R. 2019. Race after technology: Abolitionist tools for the new Jim code. John Wiley & Sons.
  10. Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence.
  11. Typology of Risks of Generative Text-to-Image Models. ArXiv:2307.05543 [cs].
  12. Expert biases in technology foresight. Why they are a problem and how to mitigate them. Technological forecasting & social change, 151. ISBN: 0040-1625.
  13. Scenario types and techniques: towards a user’s guide. Futures, 38(7): 723–739.
  14. Brey, P. 2017. Ethics of emerging technology. The ethics of technology: Methods and approaches, 175–191.
  15. Burnam-Fink, M. 2015. Creating narrative scenarios: Science fiction prototyping at Emerge. Futures, 70: 48–55.
  16. AHA!: Facilitating AI Impact Assessment by Generating Examples of Harms. ArXiv:2306.03280 [cs].
  17. Scenario types and techniques: Towards a user’s guide. Futures, 38(7): 723–739. ISBN: 0016-3287.
  18. Is Knowledge All Large Language Models Needed for Causal Reasoning? arXiv preprint arXiv:2401.00139.
  19. Omission and commission errors underlying AI failures. AI & society, 1–24.
  20. Generative adversarial networks in medical image augmentation: a review. Computers in Biology and Medicine, 144: 105382.
  21. Chatbot arena: An open platform for evaluating llms by human preference. arXiv preprint arXiv:2403.04132.
  22. Commission, E. 2024. Proposal for a Regulation of the European Parliament and of the Council laying down harmonised rules on Artificial Intelligence (Artificial Intelligence Act) and amending certain Union legislative acts, Pub. L. No. COM(2021) 206 final.
  23. How do we speak about algorithms and algorithmic media futures? Using vignettes and scenarios in a citizen council on data-driven media personalisation. New Media & Society, 14614448241232589.
  24. Anticipating and addressing the ethical implications of deepfakes in the context of elections. New media & society, 23(7): 2072–2098.
  25. Fuerth, L. 2011. Operationalizing Anticipatory Governance. PRISM, 2(4): 31–46. Publisher: Institute for National Strategic Security, National Defense University.
  26. The concept of plausibility in a risk analysis context: Review and clarifications of defining ideas and interpretations. Safety science, 147: 105635.
  27. Guston. 2014. Understanding ‘anticipatory governance’. Social Studies of Science, 44(2): 218–242.
  28. Global AI ethics: a review of the social impacts and ethical implications of artificial intelligence. arXiv preprint arXiv:1907.07892.
  29. Adding Structure to AI Harm. Technical report, Center for Security and Emerging Technology.
  30. Cladder: Assessing causal reasoning in language models. In Thirty-seventh Conference on Neural Information Processing Systems.
  31. Causal reasoning and large language models: Opening a new frontier for causality. arXiv preprint arXiv:2305.00050.
  32. Anticipating Impacts: Using Large-Scale Scenario Writing to Explore Diverse Implications of Generative AI in the News Environment. ArXiv:2310.06361 null.
  33. My Future with My Chatbot: A Scenario-Driven, User-Centric Approach to Anticipating AI Impacts.
  34. Causal Reasoning and Large Language Models: Opening a New Frontier for Causality. ArXiv:2305.00050 [cs, stat].
  35. Implicit representations of meaning in neural language models. arXiv preprint arXiv:2106.00737.
  36. Causal discovery with language models as imperfect experts. arXiv preprint arXiv:2307.02390.
  37. Can large language models build causal graphs? arXiv preprint arXiv:2303.05279.
  38. Better Call GPT, Comparing Large Language Models Against Lawyers. arXiv preprint arXiv:2401.16212.
  39. Martínez, E. 2024. Re-Evaluating GPT-4’s bar exam performance. Artificial Intelligence and Law, 1–24.
  40. Algorithmic impact assessments and accountability: the co-construction of impacts. Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency.
  41. Auditing Recommender Systems. Putting the DSA into practice wit a risk-scenario-based approach. Technical report, Stiftung Neue Verantwortung.
  42. Assembling Accountability: Algorithmic Impact Assessment for the Public Interest. SSRN Electronic Journal.
  43. Anticipatory ethics and the role of uncertainty. arXiv preprint arXiv:2011.13170.
  44. Nikolova, B. 2014. The rise and promise of participatory foresight. European Journal of Futures Research, 2(1). ISBN: 2195-4194.
  45. NIST. 2023. NIST Risk Management Framework Aims to Improve Trustworthiness of Artificial Intelligence.
  46. Interim Measures for the Management of Generative Artificial Intelligence Services.
  47. Generative agents: Interactive simulacra of human behavior. In Proceedings of the 36th Annual ACM Symposium on User Interface Software and Technology, 1–22.
  48. Sarewitz, D. 2011. Anticipatory governance of emerging technologies. The growing gap between emerging technologies and legal-ethical oversight: The pacing problem, 95–105.
  49. Selbst, A. D. 2021. An Institutional View of Algorithmnic Impact. Harvard Journal of Law & Technology, 35(1). Publisher: HeinOnline.
  50. Selin, C. 2006. Trust and the illusive force of scenarios. Futures : the journal of policy, planning and futures studies, 38(1): 1–14. ISBN: 0016-3287.
  51. Sociotechnical Harms of Algorithmic Systems: Scoping a Taxonomy for Harm Reduction. ArXiv:2210.05791 [cs].
  52. Evaluating the Social Impact of Generative AI Systems in Systems and Society. ArXiv:2306.05949 [cs].
  53. A systematic review of artificial intelligence impact assessments. Artificial Intelligence Review.
  54. Gemini: a family of highly capable multimodal models. arXiv preprint arXiv:2312.11805.
  55. Causal-discovery performance of chatgpt in the context of neuropathic pain diagnosis. arXiv preprint arXiv:2301.13819.
  56. Urueña, S. 2019. Understanding “plausibility”: A relational approach to the anticipatory heuristics of future scenarios. Futures, 111: 15 – 25.
  57. Ethical and social risks of harm from language models. arXiv preprint arXiv:2112.04359.
  58. Sociotechnical safety evaluation of generative ai systems. arXiv preprint arXiv:2310.11986.
  59. Taxonomy of risks posed by language models. In Proceedings of the 2022 ACM Conference on Fairness, Accountability, and Transparency, 214–229.
  60. Empowering LLM-based machine translation with cultural awareness. arXiv preprint arXiv:2305.14328.
User Edit Pencil Streamline Icon: https://streamlinehq.com
Authors (3)
  1. Julia Barnett (7 papers)
  2. Kimon Kieslich (12 papers)
  3. Nicholas Diakopoulos (19 papers)
Citations (1)
View on Semantic Scholar