Talking to AI Assistants Daily? Research Shows It Could Harm Your Well-being
Voice interaction with AI systems like ChatGPT may actually lead to worse psychological outcomes with prolonged daily use compared to traditional text interfaces, contrary to widespread industry assumptions about natural voice interaction being superior.
End of Miles reports on unexpected findings from a new joint study by OpenAI and MIT Media Lab that could challenge how tech companies design future AI products—and how consumers should use them.
Short-term benefits, long-term concerns
The research team, analyzing both real-world usage patterns and conducting controlled experiments with nearly 1,000 participants over four weeks, discovered a nuanced relationship between interaction mode and psychological effects.
"Voice modes were associated with better well-being when used briefly, but worse outcomes with prolonged daily use." OpenAI and MIT Media Lab Research Report
This finding directly contradicts the tech industry's push toward voice interfaces as the future of human-computer interaction. Companies like Apple, Google, Amazon, and now OpenAI have invested billions in developing more natural-sounding voice assistants, operating under the assumption that voice represents a more intuitive, frictionless way to interact with technology.
Not all voice is created equal
Interestingly, the research team found that the specific voice characteristics didn't significantly impact outcomes. The study tested various voice personalities and found no substantial difference in psychological impact between engaging and neutral voice types.
"Importantly, using a more engaging voice did not lead to more negative outcomes for users over the course of the study compared to neutral voice or text conditions." The researchers
The MIT-OpenAI collaboration instead pointed to duration and frequency of use as the critical factors. The research suggests that text-based interaction may create a natural barrier that better preserves psychological distance between humans and AI systems.
Why this matters now
As AI assistants become increasingly integrated into daily life, these findings arrive at a crucial moment. The current industry trend is heavily weighted toward more natural, voice-first AI interactions, with major tech platforms rolling out increasingly sophisticated voice capabilities.
The research team emphasizes that individual factors play a significant role in determining outcomes. Those who viewed AI as a friend or had stronger tendencies for attachment in relationships showed more negative effects from prolonged chatbot use.
"Extended daily use was also associated with worse outcomes. These correlations, while not causal, provide important directions for future research on user well-being." From the joint research publication
For everyday users, the implications are clear: voice interaction with AI might be best used sparingly rather than as a constant companion. And for developers, these findings suggest a need to reconsider assumptions about voice interfaces being universally superior to text-based interactions.
The team's work represents a significant departure from conventional wisdom in AI development, where voice capabilities have been seen primarily as technical challenges rather than potential psychological concerns. As OpenAI moves to update its Model Spec with these insights, other AI developers may need to follow suit in reconsidering how voice interaction is implemented in consumer products.
