Observed Signal · Jul 20, 2026 · Research Study · Source: t3n · Impact: 2/5 · Sentiment: Neutral
Study: People Trust Chatbots Despite Wrong Answers
Researchers from several French universities and one Italian university ran an experiment asking participants to answer visual detail questions about films, either alone or with access to a weak AI model (Step 3.5 Flash). The AI-assisted group produced fewer correct answers (9% vs. 27%) and were far less likely to admit ignorance (3% vs. 44%). AI-assisted participants were also more confident in their answers (76% vs. 30%). Valerio Capraro, a professor at the University of Milan-Bicocca, warned that reliance on AI may erode people's willingness to acknowledge the limits of their knowledge, with implications for critical thinking in future generations.
Findings highlight user overreliance on conversational AI and reduced admission of uncertainty, which is relevant to trust, misinformation risk, and the design of conversational ad formats and UX in the advertising ecosystem.
Track OpenAI Signals & Market Shifts in Real-Time
Polaris7 autonomous intelligence agents track regulatory filings, primary sources, executive changes, and deal flow 24/7. Create your free Explorer workspace to monitor these entities.
Key Takeaways & Evidence Grounding
- A multi-university research team (several French universities and one Italian university) conducted a study on human responses when allowed to use AI.
- The study used the model Step 3.5 Flash to provide answers to visual-detail film questions (chosen because that model performs poorly on such tasks).
- Participants without AI assistance answered correctly 27% of the time; participants with AI assistance answered correctly 9% of the time.
- In the no-AI group, 44% admitted they did not know the answer; in the AI-assisted group only 3% admitted ignorance.
- AI-assisted participants reported higher confidence: 76% said they were confident vs. 30% in the no-AI group.
Connected Companies & Entities
6 Entities mapped“The article refers to chatbots such as ChatGPT (a product of OpenAI) when discussing conversational AI tools....”
“The article refers to the chatbot Claude (a product by Anthropic) when discussing conversational AI tools....”
“The article refers to the chatbot/model Gemini (a Google product) when discussing conversational AI tools....”
“Article published on the German tech publisher t3n.de....”
“The article includes embedded external content from TargetVideo GmbH....”
“Valerio Capraro's comments in the article are reported as being made to The Register....”
Ontology Mapping & Concepts
Related Market Signals & Shifts
Recent verified developments and strategic activity across this market segment.
Study: Chatbots Weaken Misinformation Detection
An MIT Media Lab study found that relying on AI systems for fact‑checking over the course of a month reduces users’ independent ability to detect misinformation once the chatbot is unavailable. In a four‑week experiment with 67 participants, AI assistance improved misinformation detection by 21%, but when AI was removed performance in week four fell 15 percentage points below baseline; roughly one quarter of participants believed they had improved despite performing worse. The article also cites a separate review of 22 public broadcasters’ tests of ChatGPT, Microsoft Copilot, Google Gemini and Perplexity AI that found nearly half of AI responses had at least one significant issue (31% had major citation problems; 20% contained serious factual errors). Authors warn that conversational styles that narrate answers can create dependency, while socratic questioning may better support learning. The study notes sample limitations and plans broader follow-ups.
Study: People Overestimate AI Systems' Competence
An international research team from the University of Waterloo and University College London found that people systematically overestimate the self-confidence and competence of AI-generated answers compared with human responses, even when the behaviour shown is identical. The study, published in Communications Psychology, reports that users infer confidence from cues such as response speed and perceived system ability; AI systems like ChatGPT and Gemini typically do not communicate their own uncertainty. Clara Colombatto (University of Waterloo) warns that this can cause excessive trust in AI recommendations. The researchers are working on ways to help people better assess AI competence and to provide users with tools or signals that convey model uncertainty.
Study: Blind Trust in AI Lowers Self-Confidence
A Middlesex University study published in the journal Technology, Mind, and Behavior reports that heavy reliance on generative AI tools (examples cited: ChatGPT, Claude, Gemini) is associated with users saying the tools 'do the thinking for them', lower trust in their own reasoning, and reduced sense of ownership for ideas. The study followed 1,923 adults in the U.S. and Canada across tasks such as drafting a salary-negotiation plan and interpreting ambiguous data. Participants who actively edited, questioned or rejected AI outputs reported higher self-confidence and stronger ownership over results. The article also cites complementary survey data: McKinsey’s HR-Monitor 2026 showing regular AI use in Germany doubled from 19% to 38%, an Oxford University Press survey of 2,000 UK teenagers on AI in schooling, and a Bitkom finding that ~48% of German 14–19-year-olds think AI 'makes you dumb'.
Track Real-Time Market Signals & Shifts
Set up custom watchlists to receive automated, evidence-grounded executive digests whenever material signals or shifts occur across your tracked landscape.