USC researchers find communication gap between AI and humans - Daily Trojan

2026-06-24 · Source: artifical intelligence via Google News · Field: Technology & Digital — Artificial Intelligence & Machine Learning · Depth: Novice, short

Summary

USC researchers from the Viterbi School of Engineering, led by Mayank Kejriwal, found a significant communication gap between humans and Large Language Models (LLMs) regarding words of uncertainty. Their study, "An evaluation of estimative uncertainty in large language models," published in npj Complexity, revealed that AI models interpret terms like "likely" or "probably" with stronger probabilities than humans typically do. This discrepancy can lead to miscommunication, particularly as 90% of students use AI academically, according to a 2025 Copyleaks survey. The findings highlight challenges of overreliance on AI responses, especially in scientific contexts where precise communication of uncertainty is crucial. Researchers emphasize the need for users to double-check AI-generated results and seek clarification, while also exploring how AI's language might be adapted to cultural nuances without anthropomorphizing the models.

Key takeaway

For research scientists and students relying on Large Language Models for critical information or drafting scientific papers, you must actively verify AI-generated content. Your AI may interpret words of uncertainty, like "probably," with higher confidence than intended, leading to miscommunication. Always seek clarification from the AI and double-check its responses to ensure factual accuracy and prevent conveying incorrect impressions in your work.

Key insights

AI models interpret words of uncertainty like "likely" or "probably" with stronger probabilities than humans, creating a communication gap.

Principles

• AI models do not interpret common words like humans.
• Overreliance on AI responses can lead to miscommunication.
• AI's language may need cultural adaptation.

In practice

• Ask AI models for clarification on important questions.
• Double-check AI-generated results.
• Be aware AI is not a human speaker.

Topics

• Large Language Models
• AI Communication
• Uncertainty Interpretation
• Scientific Communication
• AI Bias
• Human-AI Interaction

Best for: AI Product Manager, AI Scientist, AI Student, Research Scientist

Related on AIssential

• How AI’s language barrier limits climate disaster responses — Western-centric AI training data causes misinterpretation of diverse everyday language, impacting critical applications like disaster response.
• Think AI ‘knows’ what it’s doing? Well, think again - Digital Journal — Journalists use anthropomorphic language for AI less frequently and more nuancedly than often assumed.
• Can AI Truly Understand Human Language? — AI models recognize language patterns and approximate meaning, but lack true human-like understanding.
• AI is teaching us to speak like bots, and it’s a problem — AI chatbots are subtly homogenizing human language, making our communication increasingly machine-like by nudging expression towards an average.
• Scientists invented a fake disease. AI told people it was real — AI models can propagate misinformation, even when explicitly labeled as fake, highlighting data validation challenges.
• ‘Probably’ doesn’t mean the same thing to your AI as it does to you — AI chatbots misinterpret words of estimative probability compared to humans, posing risks in high-stakes applications.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by artifical intelligence via Google News.