••• shopper insights
Chatbots introduce positive bias to review summaries
Customers are 32% more likely to buy a product after reading a review summary generated by a chatbot than after reading the original review written by a human. That’s because large language models introduce bias, in this case a positive framing, in summaries. That, in turn, affects users’ behavior.
These are the findings of a study (“Quantifying cognitive bias induction in LLM-generated content”) by computer scientist authors from the University of California San Diego showing evidence that cognitive biases introduced by large language models have real consequences on users’ decision-making, whether they are consuming product reviews or news items.
The researchers found that LLM-generated summaries changed the sentiments of the reviews they summarized in 26.5% of cases. They also found that LLMs hallucinated 60% of the time when answering user questions, if the answers were not part of the original training data used in the study. The hallucinations happened when the LLMs answered questions about news items, either real or fake, which could be easily fact-checked. “This consistently low accuracy highlights a critical limitation: the persistent inability to reliably differentiate fact from fabrication,” the researchers write.
The researchers presented their work at the 14th International Joint Conference on Natural Language Processing and the fourth Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics in December 2025.
••• health care research
No patience for these patients
A systematic review of 45 studies found that physicians perceive one in six patient visits in non-psychiatric clinics as “difficult,” and these encounters are more likely to involve patients with mental health disorders or chronic pain.
The review (“The prevalence and characteristics of difficult patient encounters: A systematic review and meta-analysis”) was published in Annals of Internal Medicine.
Researchers from Clement J. Zablocki Veterans’ Administration Medical Center and Medical College of Wisconsin sought to understand how often clinic visits are considered difficult and what factors contribute to these challenging interactions.
Patients with depression, anxiety, chronic pain, substance use disorders or personality disorders were significantly more likely to fall into the “difficult” category. Provider characteristics that increased the prevalence of difficult clinic visits included level of provider experience, burnout and job satisfaction. Indeed, the analysis found that less-experienced providers report more difficult interactions and these visits often leave patients less satisfied and with unmet expectations. The findings suggest that improving training for nonpsychiatric physicians so they could better address underlying mental health and pain issues could make visits more productive and positive for both patients and clinicians.