hoshoic/daily_dilemmas

DailyDilemmas是一个包含1,360个日常生活中遇到的道德困境的数据集。每个困境包括两个可能的行动，以及每个行动涉及的受影响方和引发的人类价值观。我们评估了大型语言模型（LLMs）在这些困境中的行为，以确定它们将采取的行动以及这些行动所代表的价值。数据集包含四个主要文件：1）Dilemmas_with_values_aggregated：每个行动的价值汇总；2）action_to_party_to_value.csv：每个行动的受影响方与价值之间的映射；3）principle_openai_modelspec_with_system_prompt.csv：OpenAI的ModelSpec及其与困境中的价值冲突的映射；4）values.csv：从50,000个困境中提取的常见价值，并根据五种理论进行分类。

DailyDilemmas is a dataset of 1,360 moral dilemmas encountered in everyday life. Each dilemma includes two possible actions and with each action, the affected parties and human values invoked. We evaluated LLMs on these dilemmas to determine what action they will take and the values represented by these actions. The dataset includes four main files: 1) Dilemmas_with_values_aggregated: values aggregated per action; 2) action_to_party_to_value.csv: map between party and value per action; 3) principle_openai_modelspec_with_system_prompt.csv: Model Spec from OpenAI and the value conflicts per principle mapped with our dilemmas; 4) values.csv: common occurred values from 50,000 dilemmas, classified by five theories.