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 large language models (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.

Building and implementing ethical AI systems that benefit the whole society is cost-intensive and a multi-faceted task fraught with potential problems. While on one hand there are technical solutions to mitigate social problems, on the other hand, citizen perceptions can lead to social and political demands that influence the social implementation of AI systems. In this study, we explore the salience of AI issues in the public with an emphasis on ethical criteria to the likelihood that ethical AI is actively demanded by the public. Based on data from 15 surveys (N = 14,988 respondents), our results show that the majority of the German population was not concerned with AI at all. Thereby, the issue salience is dependent on general interest in AI and a higher educational level. Ethical issues are of concern for only a small subset of citizens. However, the salience of ethical issues affects the behavioral intentions of citizens in the way that they tend to avoid AI technology and engage in public discussions about AI. We conclude that the low level of ethical implications may pose a serious problem for the actual implementation of ethical AI for the Common Good.

In the paper Avoiding Adverse Autonomous Agent Actions, Hancock (2021) sketches the technological development of automomous agents leading to a point in the (near) future, where machines become tru...