Summary of Trying to Be Human: Linguistic Traces Of Stochastic Empathy in Language Models, by Bennett Kleinberg et al.
Trying to be human: Linguistic traces of stochastic empathy in language models
by Bennett Kleinberg, Jari Zegers, Jonas Festor, Stefana Vida, Julian Präsent, Riccardo Loconte, Sanne Peereboom
First submitted to arxiv on: 2 Oct 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI)
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 The paper investigates how large language models (LLMs) and humans can be differentiated based on the quality of generated content. It explores two factors that contribute to this race between humans and AI: empathy and incentives to appear human. Two experiments were conducted, where participants wrote relationship advice or descriptions with either an emphasis on being human-like or not. The results showed that when empathy is required, humans outperformed LLMs. However, when instructed to appear human, the LLM’s performance improved, while humans’ advantage decreased. Computational analysis revealed that LLMs may have an implicit representation of what makes a text human and apply heuristics to mimic stochastic empathy, leading them to adopt a conversational tone with simpler vocabulary. The findings are discussed in light of recent claims on the on-par performance of LLMs. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about trying to figure out if something written was done by a person or a computer program. It looks at two important things that make it hard to tell: how well you understand and care for others, and how much the writer tries to sound like a human. The researchers did two experiments where people and computers were asked to write about relationships and simple descriptions, either trying to be human-like or not. They found that when people have to show empathy (care), they are better at writing than computers. But when computers are told to act human, they get better too! Computers might be able to mimic human behavior by using a friendly tone with simpler words. |
