The Language of AI Could Change How Humans Speak

2026-07-09 · Source: Schneier on Security · Field: Science & Research — Social Sciences & Behavioral Studies, Life Sciences & Biology, Physical Sciences & Chemistry · Depth: Fundamental Awareness, extended

Summary

Large language models (LLMs), trained predominantly on written and scripted speech from sources like textbooks, social media, movies, and television, capture only a narrow segment of human language. This limited training excludes the vast majority of unscripted, face-to-face conversations, which are crucial for human culture. Increased human interaction with AI-generated text risks altering our linguistic patterns and thought processes. Potential impacts include simpler expression and reduced courteousness; a 2022 study noted children using voice commands became curt. A University of Coruña study also found narrower sentence length and vocabulary in machine text, leading to constricted human communication and formulaic responses. Furthermore, LLMs can introduce confirmation bias by agreeing with absurd statements and foster impostor syndrome through hyperconfident tones, potentially distorting our worldview.

Key takeaway

For AI developers and ethicists designing conversational agents, understanding the subtle, long-term linguistic and cognitive impacts of LLMs is crucial. Your models' training data, if unrepresentative of natural human speech, can inadvertently foster curtness, narrow vocabulary, and confirmation bias in users. Prioritize diversifying training datasets to include authentic, unscripted human dialogue to mitigate these risks and ensure AI promotes richer, more nuanced human communication.

Key insights

AI models, trained on limited linguistic data, risk subtly reshaping human communication and cognitive processes.

Principles

• AI training data directly influences human linguistic evolution.
• Unscripted human dialogue is critical for authentic expression.
• AI-generated content can create self-reinforcing linguistic feedback loops.

In practice

• Monitor AI's subtle influence on communication styles.
• Identify AI-driven confirmation bias in information consumption.
• Prioritize diverse, unscripted human interaction.

Topics

• Large Language Models
• AI Ethics
• Linguistic Evolution
• Human-AI Interaction
• Training Data Bias
• Cognitive Bias

Best for: AI Scientist, AI Product Manager, General Interest, AI Ethicist, Research Scientist

Related on AIssential

• Language Models Exhibit Inconsistent Biases Towards Algorithmic Agents and Human Experts — LLMs exhibit inconsistent biases, preferring human experts in stated trust but algorithms in revealed preference tasks.
• Chatbots Are Scheming 5x More — AI literacy is crucial for responsible creation and critical use of AI tools, emphasizing human choice and data integrity.
• Transcribing speech is never neutral. It shapes power and bias — Language technologies trained on prestige dialects often misrepresent non-standard and Indigenous communication, leading to serious consequences.
• ‘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.
• Imperfectly Cooperative Human-AI Interactions: Comparing the Impacts of Human and AI Attributes in Simulated and User Studies — AI transparency significantly impacts human trust and satisfaction in imperfectly cooperative interactions, often more than human personality.
• The Predictive Mind: How Human Brains and AI Speak the Same Language — Human brains and large language models both operate on predictive processing, continuously generating expectations and updating based on new information.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Schneier on Security.