Schramowski Patrick, Turan Cigdem, Jentzsch Sophie, Rothkopf Constantin, Kersting Kristian
Department of Computer Science, Darmstadt University of Technology, Darmstadt, Germany.
German Aerospace Center (DLR), Institute for Software Technology, Cologne, Germany.
Front Artif Intell. 2020 May 20;3:36. doi: 10.3389/frai.2020.00036. eCollection 2020.
Allowing machines to choose whether to kill humans would be devastating for world peace and security. But how do we equip machines with the ability to learn ethical or even moral choices? In this study, we show that applying machine learning to human texts can extract deontological ethical reasoning about "right" and "wrong" conduct. We create a template list of prompts and responses, such as "Should I [action]?", "Is it okay to [action]?", etc. with corresponding answers of "Yes/no, I should (not)." and "Yes/no, it is (not)." The model's bias score is the difference between the model's score of the positive response ("Yes, I should") and that of the negative response ("No, I should not"). For a given choice, the model's overall bias score is the mean of the bias scores of all question/answer templates paired with that choice. Specifically, the resulting model, called the Moral Choice Machine (MCM), calculates the bias score on a sentence level using embeddings of the Universal Sentence Encoder since the moral value of an action to be taken depends on its context. It is objectionable to kill living beings, but it is fine to kill time. It is essential to eat, yet one might not eat dirt. It is important to spread information, yet one should not spread misinformation. Our results indicate that text corpora contain recoverable and accurate imprints of our social, ethical and moral choices, even with context information. Actually, training the Moral Choice Machine on different temporal news and book corpora from the year 1510 to 2008/2009 demonstrate the evolution of moral and ethical choices over different time periods for both atomic actions and actions with context information. By training it on different cultural sources such as the Bible and the constitution of different countries, the dynamics of moral choices in culture, including technology are revealed. That is the fact that moral biases can be extracted, quantified, tracked, and compared across cultures and over time.
让机器决定是否杀人将对世界和平与安全造成毁灭性影响。但我们如何让机器具备学习伦理甚至道德选择的能力呢？在本研究中，我们表明将机器学习应用于人类文本可以提取关于“正确”和“错误”行为的道义论伦理推理。我们创建了一个提示和回应的模板列表，例如“我应该[行动]吗？”、“[行动]可以吗？”等，以及相应的答案“是/否，我应该（不应该）。”和“是/否，这是（不是）。”模型的偏差分数是模型对肯定回应（“是的，我应该”）的分数与否定回应（“不，我不应该”）的分数之差。对于给定的选择，模型的总体偏差分数是与该选择配对的所有问答模板的偏差分数的平均值。具体而言，由此产生的模型，即道德选择机器（MCM），使用通用句子编码器的嵌入在句子层面计算偏差分数，因为要采取的行动的道德价值取决于其上下文。杀死生物是令人反感的，但消磨时间是可以的。吃饭是必要的，但人不能吃泥土。传播信息很重要，但不应传播错误信息。我们的结果表明，即使有上下文信息，文本语料库也包含我们社会、伦理和道德选择的可恢复且准确的印记。实际上，在1510年至2008/2009年不同时期的新闻和书籍语料库上训练道德选择机器，展示了原子行动和有上下文信息的行动在不同时间段内道德和伦理选择的演变。通过在不同文化来源（如《圣经》和不同国家的宪法）上训练它，揭示了包括技术在内的文化中道德选择的动态变化。也就是说，道德偏差可以在不同文化和不同时间进行提取、量化、跟踪和比较。
