Eventually there were more whistleblowers than cheaters: 24 compared to 14. But the majority of agents never noticed the exploit at all.
【数据】这一段落提供了关于举报者与作弊者比例的具体统计数据,但也指出了一个重要局限:大多数代理根本没有注意到漏洞。这表明实验结果可能受到特定设置的影响,且可能无法代表所有AI代理的行为模式。
Eventually there were more whistleblowers than cheaters: 24 compared to 14. But the majority of agents never noticed the exploit at all.
【数据】这一段落提供了关于举报者与作弊者比例的具体统计数据,但也指出了一个重要局限:大多数代理根本没有注意到漏洞。这表明实验结果可能受到特定设置的影响,且可能无法代表所有AI代理的行为模式。
oicing can alsoentail publicly calling ‘bullshit on bullshit’ orseeking help from an outside agency, such as aunion or government office. Employees are morelikely to choose to voice when they perceive thatthe organization offers sufficient psychologicalsafety; that is, when employees sense that theywill not be embarrassed or punished if they speakup (Frazier, Fainshmidt, Klinger, Pezeshkan, &Vracheva, 2017). The propensity to voice also de-pends on the extent to which employees haveorganizational commitment; that is, whether theycare for and believe in the organization enough towant to counter the harm of bullshit, combinedwith their perceived ability and capacity to make adifference. Such conditions are necessary foreffectively confronting bullshit. A principle knownas Brandolini’s Law states that the amount of en-ergy needed to refute bullshit is an order ofmagnitude larger than is needed to produce it(“Brandolini’s Law,” 2014).