Presentation Tae Wan Kim, March 2, 2020:
3:30 PM Hermann Hall Lounge
Title: When Is It Not Unethical For AI to Lie to Humans? Trust-based Theory and Empirical Evidence
As businesses increasingly utilize artificial intelligence (AI) to make important decisions for humans, societal worries about the compatibility of AI and human values grow. In response, researchers have examined how to make AI learn ethical principles as well as calculative and strategic intelligence. Such attempts are lumped under the broader term "value alignment.” In this project, we attempt to ground conversational AI agents' behavior in ethical values. In particular, we clarify under what conditions it is not unethical for AI to lie to human users. In Section 1 we develop a framework of when it is unethical for AI to lie, using cases of a negotiation chatbot, which we find distinctively useful to clarify the ethics of AI lies. In Section 2, we explain why the ethical status of conversational AI in part depends upon cultural norms, using survey data from various countries including US, China, Netherlands and Sweden. By doing so, we show that a particular kind of AI lie that is unethical in a certain culture is not necessarily unethical in a different culture. Finally we discuss implications for AI companies’s ethical duties on how to program AI and how to develop a fully automated negotiation AI that is able to identify under what conditions it is ethically okay to lie.
Tae Wan Kim is Associate Professor of Business Ethics at Carnegie Mellon’s Tepper School of Business. Kim is a recipient of Business Ethics Quarterly's Best Article Award (2015, 2017) and was selected as one of 11 Groundbreaking, World-Changing Wharton PhDs in 2018. Kim has served on editorial boards of Business Ethics Quarterly and Journal of Business Ethics. Kim has served a program committee of AAAI/ACM Conference on Artificial Intelligence, Society and Ethics (AIES) and is a committee member of the IEEE Global Initiative for Ethical Considerations in Artificial Intelligence and Autonomy Systems. Kim has taught ethics to MBA students at Carnegie Mellon University since 2012. Kim has also severed as speakers on AI ethics for public events such as TTI Vanguard and MIT Robo-AI Exchange.