被算法腐蚀？人工智能生成与人类撰写的建议如何塑造（不）诚实行为

中文摘要

人工智能正日益成为不可或缺的顾问。若人工智能劝说人们做出不诚实行为，便会引发新的伦理关切。在一项实验中，我们研究人工智能建议（由自然语言处理算法生成）如何影响（不）诚实行为，将其与等价的人类建议相比较，并检验就建议来源保持透明是否重要。我们发现，鼓励不诚实的建议会增加不诚实行为，而鼓励诚实的建议并不会增加诚实行为。人工智能建议与人类建议均是如此。算法透明——一项常被提出用以缓解人工智能风险的政策——并不影响行为。这些发现标志着朝着负责任地管理人工智能建议迈出的第一步。

Abstract

Artificial intelligence increasingly becomes an indispensable advisor. New ethical concerns arise if artificial intelligence persuades people to behave dishonestly. In an experiment, we study how artificial intelligence advice (generated by a natural language processing algorithm) affects (dis)honesty, compare it to equivalent human advice and test whether transparency about the advice source matters. We find that dishonesty-promoting advice increases dishonesty, whereas honesty-promoting advice does not increase honesty. This is the case for both artificial intelligence and human advice. Algorithmic transparency, a commonly proposed policy to mitigate artificial intelligence risks, does not affect behaviour. The findings mark the first steps towards managing artificial intelligence advice responsibly.