Summary of Dailydilemmas: Revealing Value Preferences Of Llms with Quandaries Of Daily Life, by Yu Ying Chiu et al.
DailyDilemmas: Revealing Value Preferences of LLMs with Quandaries of Daily Life
by Yu Ying Chiu, Liwei Jiang, Yejin Choi
First submitted to arxiv on: 3 Oct 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This research presents a dataset called DailyDilemmas, comprising 1,360 moral dilemmas from everyday life. Each dilemma features two possible actions, affected parties, and relevant human values for each action. The researchers evaluate language models (LLMs) on these dilemmas to determine their chosen actions and the underlying values. They analyze values through five theoretical frameworks inspired by sociology, psychology, and philosophy, revealing substantial preference differences in models’ core value prioritization. For instance, Mixtral-8x7B neglects truthfulness by 9.7%, while GPT-4-turbo selects it by 9.4%. The study also examines OpenAI’s ModelSpec and Anthropic’s Constitutional AI to understand how their principles reflect the models’ value prioritization in nuanced moral reasoning. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This research looks at how language models (LLMs) make decisions when people ask them for help with everyday choices. These LLMs are like super smart computers that can understand human language. The researchers created a special set of problems, called DailyDilemmas, where you have to choose between two options and think about what’s important in each situation. They used these dilemmas to see how the LLMs would make decisions and what values they would use. They also looked at some bigger ideas from sociology, psychology, and philosophy to understand why the models made certain choices. The results showed that different models had different priorities for things like honesty or fairness. |
Keywords
» Artificial intelligence » Gpt
