AI overestimates how smart people are, according to HSE economists
National Research University Higher School of Economics
Scientists at HSE University have found that current AI models, including ChatGPT and Claude, tend to overestimate the rationality of their human opponents—whether first-year undergraduate students or experienced scientists—in strategic thinking games, such as the Keynesian beauty contest. While these models attempt to predict human behaviour, they often end up playing 'too smart' and losing because they assume a higher level of logic in people than is actually present. The study has been published in the Journal of Economic Behavior & Organization.
In the 1930s, British economist John Maynard Keynes developed the theoretical concept of a metaphorical beauty contest. A classic example involves newspaper readers being asked to select the six most attractive faces from a set of 100 photos. The prize is awarded to the participant whose choices are closest to the most popular selection—that is, the average of everyone else’s picks. Typically, people tend to choose the photos they personally find most attractive. However, they often lose, because the actual task is to predict which faces the majority of respondents will consider attractive. A rational participant, therefore, should base their choices on other people’s perceptions of beauty. Such experiments test the ability to reason across multiple levels: how others think, how rational they are, and how deeply they are likely to anticipate others’ reasoning.
Dmitry Dagaev, Head of the Laboratory of Sports Studies at the Faculty of Economic Sciences, together with colleagues Sofia Paklina and Petr Parshakov from HSE University–Perm and Iuliia Alekseenko from the University of Lausanne, Switzerland, set out to investigate how five of the most popular AI models—including ChatGPT-4o and Claude-Sonnet-4—would perform in such an experiment. The chatbots were instructed to play Guess the Number, one of the most well-known variations of the Keynesian beauty contest.
According to the rules, all participants simultaneously and independently choose a number between 0 and 100. The winner is the one whose number is closest to half (or two-thirds, depending on the experiment) of the average of all participants’ choices. In this contest, more experienced players attempt to anticipate the behaviour of others in order to select the optimal number. To investigate how a large language model (LLM) would perform in the game, the authors replicated the results of 16 classic Guess the Number experiments previously conducted with human participants by other researchers. For each round, the LLMs were given a prompt explaining the rules of the game and a description of their opponents—ranging from first-year economics undergraduates and academic conference participants to individuals with analytical or intuitive thinking, as well as those experiencing emotions such as anger or sadness. The LLM was then asked to choose a number and explain its reasoning.
The study found that LLMs adjusted their choices based on the social, professional, and age characteristics of their opponents, as well as the latter’s knowledge of game theory and cognitive abilities. For example, when playing against participants of game theory conferences, the LLM tended to choose a number close to 0, reflecting the choices that typically win in such a setting. In contrast, when playing against first-year undergraduates, the LLM expected less experienced players and selected a significantly higher number.
The authors found that LLMs are able to adapt effectively to opponents with varying levels of sophistication, and their responses also displayed elements of strategic thinking. However, the LLMs were unable to identify a dominant strategy in a two-player game.
The Keynesian beauty contest has long been used to explain price fluctuations in financial markets: brokers do not base their decisions on what they personally would buy, but on how they expect other market participants to value a stock. The same principle applies here—success depends on the ability to anticipate the preferences of others.
'We are now at a stage where AI models are beginning to replace humans in many operations, enabling greater economic efficiency in business processes. However, in decision-making tasks, it is often important to ensure that LLMs behave in a human-like manner. As a result, there is a growing number of contexts in which AI behaviour is compared with human behaviour. This area of research is expected to develop rapidly in the near future,' Dagaev emphasised.
The study was conducted with support from HSE University's Basic Research Programme.
Journal
Journal of Economic Behavior & Organization
Article Title
Strategizing with AI: Insights from a beauty contest experiment
Article Publication Date
21-Dec-2025
Survey reveals ethical gaps slowing AI
adoption in pediatric surgery
Zhejiang University
image:
Ethical concerns in the use of artificial intelligence (AI) in pediatric surgery practice among study participants.
view moreCredit: World Journal of Pediatric Surgery (WJPS)
Artificial intelligence (AI) is rapidly advancing across modern healthcare, yet its role in pediatric surgery remains limited and ethically complex. This study reveals that although surgeons recognize AI’s potential to enhance diagnostic precision, streamline planning, and support clinical decision-making, its practical use is still rare and mostly academic. Pediatric surgeons expressed strong concerns about accountability in the event of AI-related harm, the difficulty of obtaining informed consent for children, the risk of data privacy breaches, and the possibility of algorithmic bias. By examining pediatric surgeons’ experiences and perceptions, this study highlights the critical barriers that must be addressed before AI can be safely and responsibly integrated into pediatric surgical care.
At present, throughout the world, AI is reshaping how medical data are interpreted, how risks are predicted, and how complex decisions are supported. Yet pediatric surgery faces unique ethical challenges due to children’s limited autonomy, the need for parental decision-making, and the heightened sensitivity of surgical risks. In low-resource settings, concerns about infrastructure, data representativeness, and regulatory preparedness further complicate adoption. Pediatric surgeons must balance innovation with the obligation to protect vulnerable patients and maintain trust. These pressures intensify debates around transparency, fairness, and responsibility in the use of AI tools. It was with these challenges that a deeper research is needed to guide the ethical and practical integration of AI in pediatric surgical care.
A national team of pediatric surgeons from the Federal Medical Centre in Umuahia, Nigeria, has released the first comprehensive survey examining how clinicians perceive the ethical and practical implications of integrating AI into pediatric surgical care. Published (DOI: 10.1136/wjps-2025-001089) on 20 October 2025 in the World Journal of Pediatric Surgery (WJPS), the study gathered responses from surgeons across all six geopolitical zones to assess levels of AI awareness, patterns of use, and key ethical concerns. The findings reveal a profession cautiously weighing AI’s potential benefits against unresolved questions regarding accountability, informed consent, data privacy, and regulatory readiness.
The study analyzed responses from 88 pediatric surgeons, most of whom were experienced consultants actively practicing across diverse clinical settings. Despite global momentum in AI-enabled surgical innovation, only one-third of respondents had ever used AI, and their use was largely restricted to tasks such as literature searches and documentation rather than clinical applications. Very few reported using AI for diagnostic support, imaging interpretation, or surgical simulation, highlighting a substantial gap between emerging technological capabilities and everyday pediatric surgical practice.
Ethical concerns were nearly universal. Surgeons identified accountability for AI-related errors, the complexity of securing informed consent from parents or guardians, and the vulnerability of patient data as major sources of hesitation. Concerns also extended to algorithmic bias, reduced human oversight, and unclear legal responsibilities in the event of harm. Opinions on transparency with families were divided. While many supported informing parents about AI involvement, others felt disclosure was unnecessary when AI did not directly influence clinical decisions.
Most respondents expressed low confidence in existing legal frameworks governing AI use in healthcare. Many called for stronger regulatory leadership, clearer guidelines, and standardized training to prepare pediatric surgeons for future AI integration. Collectively, the findings underscore an urgent need for structured governance and capacity building.
“The results show that pediatric surgeons are not opposed to AI—they simply want to ensure it is safe, fair, and well regulated,” the research team explained. “Ethical challenges such as accountability, informed consent, and data protection must be addressed before clinicians can confidently rely on AI in settings involving vulnerable children. Clear national guidelines, practical training programs, and transparent standards are essential to ensure that AI becomes a supportive tool rather than a source of uncertainty in pediatric surgical care.”
The study underscores the need for pediatric-specific ethical frameworks, clearer consent procedures, and well-defined accountability mechanisms for AI-assisted care. Strengthening data governance, improving digital infrastructure, and expanding AI literacy among clinicians and families will be essential for building trust. As AI continues to enter surgical practice, these measures offer a practical roadmap for integrating innovation while safeguarding child safety and public confidence.
###
References
DOI
Original Source URL
https://doi.org/10.1136/wjps-2025-001089
About World Journal of Pediatric Surgery
World Journal of Pediatric Surgery (WJPS), founded in 2018, is the open-access, peer-reviewed journal in pediatric surgery area. Sponsored by Zhejiang University and Children’s Hospital, Zhejiang University School of Medicine, and published by BMJ Group. WJPS aims to be a leading international platform for advances in pediatric surgical research and practice. Indexed in PubMed, ESCI, Scopus, CAS, DOAJ, and CSCD, WJPS achieved the latest impact factor (IF) of 1.3/Q3, CiteScore of 1.5, and an estimate 2025 IF of approximately 2.0.
Journal
World Journal of Pediatric Surgery
Subject of Research
Not applicable
Article Title
Ethical considerations and challenges in the use of artificial intelligence in pediatric surgical practice: a national survey of Nigerian pediatric surgeons
ChatGPT Atlas is the worst browser for
privacy in 2025
By Dr. Tim Sandle
SCIENCE EDITOR
December 24, 2025
ChatGPT was touted as 'the heart' of OpenAI's new Atlas browser, as tech firms seek to challenge Google Chrome's supremacy - Copyright AFP/File Kirill KUDRYAVTSEV
ChatGPT Atlas failed 100% of state partitioning tests, meaning it cannot prevent websites from tracking users across different browsing sessions and Chrome ranks among the worst for privacy despite its popularity, scoring only 41 out of 100 for securing how it connects you to websites.
These are findings that appear in a browser privacy report from the company Digitain, who examined 13 popular browsers across various security features. Each browser underwent dozens of technical tests assessing three key areas: how easily websites can track or identify users (fingerprinting and tracking), how well the browser blocks data collection through cookies and trackers, and whether it maintains secure connections between websites. Scores from each category were combined into a final Privacy Risk Score ranging from 1 to 99, with higher scores indicating poorer privacy protection, and 99 representing almost no privacy at all.
The 10 browsers with the highest privacy risks
| Browser | Privacy & Anti-Fingerprinting Index Score | Tracker & Data Blocking Index Score | Connection & Navigation Security Index Score | Privacy Risk Score |
| ChatGPT Atlas | 1 | 0 | 24 | 99 |
| Chrome | 68 | 0 | 41 | 76 |
| Vivaldi | 68 | 0 | 37 | 75 |
| Edge | 68 | 15 | 41 | 63 |
| Opera | 68 | 25 | 41 | 58 |
| Ungoogled | 71 | 25 | 41 | 55 |
| Firefox | 74 | 25 | 51 | 50 |
| Safari | 80 | 25 | 47 | 49 |
| DuckDuckGo | 58 | 79 | 51 | 44 |
| Tor | 52 | 70 | 83 | 40 |
Chrome comes second with a privacy risk score of 76, despite being the world’s most popular browser. Google’s browser scored 68 in stopping websites from identifying you but received 0 points for blocking trackers, meaning companies can freely follow your activity across the web. Chrome doesn’t block tracking cookies or the hidden markers that websites use to monitor where you go online. The browser earned 41 points for connection security, falling well below privacy-focused alternatives.
Vivaldi ranks third among the least private web clients. The browser matched Chrome’s 68-point anti-fingerprinting score but similarly failed to block any trackers, recording 0 in that category. Vivaldi’s connection security also came in slightly lower at 37 points compared to Chrome’s 41. Despite marketing itself as a privacy-conscious alternative to mainstream browsers, Vivaldi’s test results show it offers low protection against online tracking and data harvesting.
Next comes Microsoft Edge. The browser posted the same 68 points as Chrome and Vivaldi for anti-fingerprinting but showed slight improvement in tracker blocking with a score of 15 out of 100. Edge tied Chrome’s connection security rating at 41 points. While Edge performs marginally better than Chrome and Vivaldi, it still fails to provide adequate privacy protection for users concerned about data collection and online tracking.
Opera rounds out the five least private browsers. Like other Chromium-based browsers in this ranking, Opera scored 68 for anti-fingerprinting and 41 for connection security. However, Opera showed slightly better tracker blocking capabilities with 25 points, putting it ahead of Edge, Vivaldi, and Chrome in this category. Despite these small improvements, Opera still ranks among browsers that offer poor overall privacy protection.
While AI-powered browsers like ChatGPT Atlas and Comet from Perplexity are getting considerable attention they do pose security issues. Since AI works by collecting and learning from data, these tools might be gathering more of personal information than users realise.
