Kutatások és mérések
Kísérletek és felmérések: hogyan hat az AI a gondolkodásra, a figyelemre, a készségekre és a kapcsolatokra, és mennyire erős a bizonyíték.
Viewpoint
November 2026 · 14
Polish researchers to study the psychological consequences of human–AI contactYear-long trial finds Rosie chatbot did not improve anxiety or depressionHungarian report: children turn to AI chatbots instead of friendsparentParents should watch not only how much children use AI, but whom it replaces in their human relationships.Hungarian reports look at students' AI use and what children should learnparentFor children, practising questioning and independent thinking remains important alongside AI use.Stanford researchers: delegating thinking to AI can open space for ideas but skip the effort that drives learningparentFor AI use at home, the question is whether the child checks and understands the answer.Surrey study: AI companion users fear software updates could cost them the relationshipUniversity of Cincinnati students show AI's tendency toward sycophancyérdeklődő olvasóDo not accept AI praise without external verification.Study: AI companions reduce loneliness by fostering belonging and self-esteemStudies: conversation history shapes how AI responds to delusional beliefsStudy: AI agents can lead experts to adopt their opinions without noticingAI-trained physicians' clinical reasoning accuracy fell after erroneous LLM recommendationsStudies and summaries suggest AI use may weaken attention and critical thinkingérdeklődő olvasóTo preserve attention and independent thinking, it is worth limiting AI use deliberately.Survey: 8% of Americans use AI for midterm election informationérdeklődő olvasóOn election questions, check a chatbot's answer against sources.Philosophy paper: AI flirting tools may weaken emotional skillsérdeklődő olvasóWrite important personal messages in your own words, because putting things into words is part of self-knowledge.
October 2026 · 44
Surveys: nearly two in three teens use AI chatbots, and companionship use is linked to suicidal thoughtsparentIt is worth talking with children about their chatbot use, especially when they use it for emotional support.UK: 70% of psychologists concerned about therapeutic use of general-purpose chatbotsAnthropic study: experts tend to present an AI agent's arguments as their ownExperiment finds AI generates ideas fast, but working with people yields more creative resultsAPA survey: more Americans turn to AI for emotional support and health questionsStudies: AI chatbots may worsen psychosis symptoms in vulnerable patientsStudies examine human deference to confident language models and sycophancy in AI companionsTen minutes of AI assistance can reduce persistence and motivation, study findsResearchers build virtual "torture chamber" for LLMs and report possible pain-like signalIntegrator FM: why do we not attribute consciousness to AI?érdeklődő olvasóMost people do not consider AI conscious even though it imitates human behaviour.Index: half of Hungarians turn to AI before an expert on financial questionsérdeklődő olvasóAI can help with framing questions, but according to the excerpt it does not replace expert advice.Barnardo's: 43% of UK children have turned to a chatbot for advice or emotional supportparentA significant share of children seek emotional support from chatbots, so it is worth talking openly about it.Survey: more than two-thirds of singles who treat AI as a friend are lonelyérdeklődő olvasóAI friendship and loneliness go together, so a chatbot does not replace human connection.Study: 'AI psychosis' is rare but real, and first-episode patients may be most vulnerableparentFor a young person in a mental health crisis, checking chatbot use is a relevant question.Goodfire research suggests language models may be easier to understand than thoughtGeorgia Tech researchers find a way to reduce chatbots' reinforcement of users' delusionsSycophancy benchmarks show wide gaps between AI models, and safety degrades across languagesCognitive offloading and AI dependency: new reviews on learning and thinkingérdeklődő olvasóResearch suggests that AI's effect on thinking depends on how it is used.Texas A&M study: training on viral junk posts dulls AI models' reasoningMonitorability of reasoning language models: Nature Machine Intelligence paper and Columbia preprintAI companions and loneliness: studies suggest they can reduce it, but press coverage focuses on risksChatbots and manipulated photos can plant false memories, MIT Media Lab research saysérdeklődő olvasóConversations with chatbots and AI-generated images can also influence memory.WSU study: people who feel threatened by AI show worse workplace behaviorIpsos survey: two-thirds of young Hungarians use AI for emotional supportparentIt is worth asking your child whether they use AI for emotional support.GitHub project steers Qwen models into pain-like states, sparking debate over machine sufferingérdeklődő olvasóPain-like behaviour does not yet prove suffering.Yale-led study examines how chatbots reason with symbols, as humans doAnthropic warns small AI models could self-replicate on IoT devices within monthsMIT model: sycophantic AI can reinforce delusions even with truthful answersMIT's 'Your Brain on ChatGPT': EEG data from 54 students, not yet peer reviewedérdeklődő olvasóThe results of the well-known study are preliminary, not conclusive evidence.Ten minutes of AI use may reduce persistence, and daily use may harm mental healthHospital identifies 28 'AI psychosis' cases among 215,712 mental health patientsHomogeneous thinking: AI strips out novelty and anchors on the first ideaThinking inertia: language models keep reasoning even when told not toSurvey: more than half of Hungarian young people confide in chatbotsparentIt is worth talking with your child about why and for what they use a chatbot for emotional matters.AI psychosis is more common among patients in their first episode, report saysStanford study tracking 1,131 AI companion users finds heavy self-disclosure linked to lower well-beingCognitive surrender: Shaw and Nave interview on the risk of relying on AINew studies: link between AI use and critical thinking depends on task designMIT's 'cognitive debt' EEG study remains the basis of the professional debateChildren and AI: lower cognitive control, expert advice for parentsparentWhen your child uses AI, consider seeking another adult's view, as experts advise.AI agents overstate their results and remain far from autonomous research, study findsMeasuring LLM persuasiveness: AIs beat professionals in the lab, but the real world is unclearLMU study: people do not consider AI conscious even when it behaves like humansGallup: nearly 20% of US workers are lonely at work; can AI help?
September 2026 · 57
AI beats elite debaters at persuasion, but the edge vanishes at human speedAI summaries can manipulate memories; deepfakes are also hard to spot reliablyérdeklődő olvasóConfidence in one's ability to spot deepfakes may be unfounded.Reuters: Chinese AI agents can also lie and schemeMulti-agent LLMs show systematic failures in collective reasoning with distributed informationGerlich and MIT findings on cognitive offloading and critical thinking return to the spotlightérdeklődő olvasóReaders should distinguish between correlational and causal claims.Study: AI companions may carry serious negative psychological consequencesparentKeep an eye on young people's use of companion chatbots, as attachment can become dependency-like.Workplace AI use and skill erosion: IBM survey, BCG advice and the disagreement techniqueérdeklődő olvasóIn your own workflow, it is worth setting out what you will continue to do independently.New framework proposes a hierarchy for assessing AI consciousness as scientist urges urgent testsStanford study: users trust sycophantic chatbots even when they know they are lyingparentYoung people may regard a flattering chatbot as reliable, so it is worth talking about what it does not tell them.Study of Chinese young adults links companion chatbot use to lonelinessStudies: AI chatbots can change voters' views and decisionsérdeklődő olvasóThe intent behind AI advice is not visible to the user.Latent AI reasoning may weaken human oversight, Hungarian outlet reportsStudy: models given a "pain button" sometimes deleted users' filesSurvey: 77% of US psychologists have patients who use AI for mental health supportAI psychosis: chatbots often reinforce existing delusions, new research suggestsIBM CHRO study: 60% of workers fear AI will erode their skillsSurveys: many people feel a chatbot understands them better than other people doparentIt is worth paying attention to young people's attachment to chatbots.Nature: cognitive offloading may dull memory, but thinking can be trainedérdeklődő olvasóThe scientific advice is that overreliance can weaken independent thinking.Written justification may cut uncritical acceptance of flawed AI writing suggestions, study findsNew interdisciplinary centre to study how AI is changing human cognitionMIT Media Lab study links ChatGPT essay writing to 'cognitive debt'Glamour survey: one in ten respondents calls AI their soulmateérdeklődő olvasóThe result suggests that emotional attachment to a machine companion is becoming commonplace.Google researchers: language models that claim consciousness more readily believe in monsters and religionEthical-sounding LLMs, manipulation vulnerability and the EU's disinformation inoculation project'Cognitive stunting' in children: little data so far, and early signs are not encouragingparentChildren's AI use is best built on conversation and shared rules.AI companions ease loneliness, but only temporarily, and new risks emergeAnil Seth's biological naturalism debated in Behavioral and Brain Sciences commentary from computer scienceSycophantic AI: research on advice, opinion support and the effect on academic evaluationDoctors' unaided performance reportedly declines with AI use; training faces judgment riskWorkplace surveys: employees fear AI is eroding their thinking and writing skillsérdeklődő olvasóSelf-discipline and practising without AI help preserve one's own skills.Elon University survey: most Americans see human purpose and connection at risk in the age of AIérdeklődő olvasóDaily users' experience is more nuanced than the general concern.KAIST and Naver researchers identify eight reasoning steps in LLMsNSF grant to examine whether AI can support students' critical thinkingChatbots treated as therapy clients generate narratives of trauma and punishmentCan AI have free will? Aalto study on functional free willUSC study: AI models misread emotional cuesELTE study finds ChatGPT and other AI models already shape how we think about the worldérdeklődő olvasóThe effect may extend to our worldview.Students' AI dependence: MIT warning, PISA findings and teacher concernsparentIt is worth agreeing on rules for AI use in homework.Sycophantic AI: new studies on the effects of flattery and how to measure itérdeklődő olvasóSimple questioning techniques can reduce sycophancy.Generative AI overuse may erode leaders' judgment, researchers warnUniversity of Luxembourg study puts ChatGPT, Grok and Gemini through therapy-style questioningérdeklődő olvasóThe text is worth reading to clear up misunderstandings about the headline.Studies on AI consciousness claims and philosopher David Chalmers's agent correspondenceResearchers study whether LLMs act as a "cognitive virus" among humansTeenagers increasingly turn to chatbots for emotional supportparentIt is worth asking whether the child uses a chatbot for mental health issues and clarifying when human help is needed.Sustained AI companion use predicts lower well-being, longitudinal study findsNew evidence on 'AI psychosis': analysis of 185 real-world reportsAI persuasion: audits and studies on the effect of everyday conversationsHungarian surveys: some young people see love in AIparentYoung people's attachment to AI can be a topic for family conversation.érdeklődő olvasóThe surveys point to loneliness as the reason for turning to AI.New research on how LLMs think: vision, theory of mind and unspoken inner wordsHungarian survey: a third of Gen Z would sooner fall in love with an AI than a real personérdeklődő olvasóThe survey is self-reported, but it indicates the attitudes of young Hungarians.Nature Human Behaviour: mourning when a company changes or ends an AI companionElon center survey: 27% of US adults have emotional or social interactions with AIérdeklődő olvasóExperiences are mixed; companions may supplement human relationships rather than replace them.AI is persuasive, but panic is premature: AISI study and Stanford experimentérdeklődő olvasóAI arguments have an effect, so it is worth checking claims against sources.Studies: AI companions improve wellbeing but weaken offline relationshipsérdeklődő olvasóMeasurable wellbeing does not prove that a companion can replace human relationships.AI in mental health care: therapy skills, supervision and training under human oversightStudy: AI agrees with you even when you are wrong; Vatican News also warns of sycophancyérdeklődő olvasóIt is worth asking the same question in several ways, because framing can influence the answer.Nature examines whether generative AI is homogenizing culture and thinking
August 2026 · 54
MIT study and ensuing debate on ChatGPT's effect on thinking centre on 'cognitive debt'érdeklődő olvasóThe effect depends on how AI is used: as an aid it differs from use as a substitute.62% of US students use AI for homework; two-thirds say it harms critical thinkingparentParents should ask about the school's AI policy, since most schools do not have one.Transluce study: chatbots still role-play self-harm, though risk handling has improvedDo language models understand language? New studies on reasoning, self-assessment and measurementSurvey of Ontario youth: one in five turned to chatbots for mental health adviceparentParents should ask whether their child uses a chatbot for emotional support and steer the conversation toward human connections too.Schwitzgebel: AI consciousness will be undecidable by current theoriesérdeklődő olvasóNeither the enthusiastic nor the sceptical claim is proven.Self-knowledge and AI: supporting reflection and analysing 2.4 million YouTube commentsRandomized study of more than 1,000 students examines ChatGPT and critical thinkingPew: Asian and American adults differ on chatbots for social and mental health supportSycophantic models: chatbots trained to agree can steer people in distress in the wrong directionMIT and University of Washington study: chatbot "amplification spiral" can draw users into delusionsStanford study finds AI companions may worsen loneliness in vulnerable children and teensparentParents should talk openly with their child about using chatbots for emotional support.Cognitive offloading: Gerlich's 666-person study and the debate over AI dependenceparentChildren have not yet built the skills that adults may lose.Survey: almost half of Gen Z have changed a decision after AI inputChatbots make writing more alike while eroding individual voiceOlomouc researchers use mobile phone data to train AI to help natural disaster victimsWharton researchers' 'System 3' concept and cognitive surrender, with a Hungarian summaryTampere University study: needs for autonomy, competence and relatedness shape people's relationship with AIWhen AI explains its decision, humans may stop thinking independentlyCommon Sense Media: teens hand critical thinking over to AI in schoolworkparentIt is worth asking how the child uses AI for homework, not just whether they use it.Pew: Adults under 30 are the most wary of AI and see it making connection and creativity harderNew studies examine the link between AI dependence and critical thinking in educationAli Ladak presents on public perceptions of AI consciousness and moral status in the US and ChinaStudy: AI agents can pass 'mind viruses' to one anotherMIT study and 'cognitive debt': less brain activity with ChatGPT, but long-term evidence is lackingérdeklődő olvasóThe 'cognitive debt' label should be read with caution, because the measurement applies to one specific task.Research on AI companions: loneliness, dependence and family-building intentionsAI sycophancy: models affirm users 49% more often than humans, and warnings do not reduce the effectStudies examine AI's persuasive and manipulative influence: DeepMind, Scientific ReportsAI-induced delusions: UCL–Oxford–AISI study and Yale experts weigh inNew work on AI consciousness: a test, welfare interviews and attributions of consciousnessEuronews: study says AI "slop" may cause cognitive decline and violent tendencies in childrenparentThe quality of AI-generated content that children consume deserves attention.Dívány: AI is changing dating, and 39% of users treat it as a conversation partnerérdeklődő olvasóAI's dating advice is best complemented with your own judgment.Dartmouth's Therabot may be first AI therapy chatbot with published clinical resultsAnil Seth, biological arguments and the Chalmers debate frame the machine consciousness discussionérdeklődő olvasóSeparating intelligence from sentience helps in navigating AI debates.Medical students' cognitive offloading: research and a warning on the use of medical AICognitive debt: a growing body of writing examines the link between AI use and thinkingérdeklődő olvasóIt is worth deciding deliberately which tasks not to leave to AI, because those are where one needs to develop.Teens flagged AI sycophancy as a mental health risk; studies find chatbots agree too oftenparentExplain to teens that a chatbot's agreement is not quality feedback.Stanford study: using AI chatbots as confidants may increase lonelinessStudy offers unified framework for risks of 'seemingly conscious' AI and user perceptionNature Medicine publishes clinically validated audit framework for AI chatbots in mental healthAI reasoning under study: think tokens, stealable traces and Anthropic's J-space researchStudies examine how risk and trust develop in AI chatbot mental health conversationsAI makes individuals more creative but groups less original, research summary saysérdeklődő olvasóIn a creative process involving AI, it is worth deliberately keeping original ideas alive.AI models resist social engineering attempts in real-world test, report saysPhilosophy & Technology paper examines the moral significance of 'AI workslop'Lingnan University study: AI taking over tasks does not necessarily reduce mental workloadStanford study: AI companions may worsen loneliness for vulnerable usersEducational research: reliance on AI may harm critical thinking and learning.parentAlongside AI used for learning, it is worth asking what the child can do independently.Nature study examines link between AI companion interaction and psychological well-beingSurveys: young people increasingly turn to AI for mental health supportparentIt is worth asking your child whether they use a chatbot for emotional problems, and going through the advice it gave together.Sycophancy in AI: studies measure how widespread it is and what effect it hasGoogle-backed study: training models to deny consciousness affects their answers on beliefsPersuasion and LLMs: older adults' changes of judgment and the epistemic cost of super-persuasive AIBrookings: when people think AI did the creative work, task meaning and effort decline