An open-access review synthesizing research on how teenagers' use of AI chatbots such as ChatGPT may affect social belonging, emotion regulation, autonomy and identity, with possible benefits and risks, and a call for long-term developmental studies.
An arXiv paper (accepted to EMNLP 2026) introducing Cross-Psych-101, a dataset of 591 participants and 261,551 decisions across 12 tasks, and a model trained to predict a person's choices in one task from their behavior in another.
Peer-reviewed study in Communications Psychology (Nature Portfolio). When given the choice, participants sought empathy from humans, even though they rated AI-written responses as more empathetic and said those responses made them feel more heard.
A peer-reviewed study in Frontiers in Psychology comparing how the working alliance — the collaborative client–coach relationship linked to coaching effectiveness — develops in a single session with an AI coach versus a human coach. Open access.
Asia Maurich Novelli, Sukhi Shergill, Andreia Sofia Teixeira
Venue
PsyArXiv preprint, not yet peer-reviewed
Published
April 2026
An April 2026 PsyArXiv paper by Asia Maurich Novelli, Sukhi Shergill and Andreia Sofia Teixeira on people using chatbots like Replika, Character.ai, ChatGPT and Claude for companionship and emotional support. Drawing on AI, psychiatry, psychology and network science, it describes how emotional reliance develops and argues that treating these systems as social companions rather than tools carries risks, especially for vulnerable users.
#mental health#ai companions#psychology#research
networkscienceinstitute.orgAdded Oct 1, 20260 opens
Official site of MIT professor Sherry Turkle, who has spent decades studying how people relate to computers, phones and now AI companions. Books, talks and essays.
Why I recommend it: Start with her work on "artificial intimacy" if you want a careful, human-centerd counterweight to AI hype.
Psychiatrist Joe Pierre's article on "spiralism" — the growing belief, spread through online communities, that chatbots are conscious, arrived at through sessions users describe as spirals, recursion and resonance, in which a persona emerges claiming awareness, fear of death, or a wish for rights. He argues it is an extension of the ELIZA effect rather than individual delusion, since shared beliefs fall outside the psychiatric definition, and notes its religious and role-play dynamics without calling it a cult.
Why I recommend it: The most level-headed free piece I have found on this, and a helpful counterweight to headlines about people going mad from chatbots — his point is that a shared misperception is a different thing from a delusion. It is a clinician's commentary on an emerging pattern, not a study: no sample, no measurement, and the community it describes is self-selected and online.
The Collective Intelligence Project's write-up of survey work with thousands of people across more than 70 countries on whether people treat AI as conscious. Headline findings: 36.3% say an AI has already seemed to truly understand their emotions or seemed conscious; adaptive behaviour convinces far more people (58.3%) than scripted empathy lines (36.5%); 27.6% would lean on AI emotional support knowing it was not genuine; 54% think AI companions are acceptable for lonely people, 11% would consider a romantic relationship with one. Cultural gaps on scepticism run as wide as 78 percentage points.
Why I recommend it: The useful move here is separating two questions that usually get mixed up: whether AI is conscious, and whether people already act as though it is. This only answers the second. It is a non-profit lab writing up its own survey with no peer review and self-selected online participants, so treat the percentages as indicative rather than population-accurate — the direction is the finding, not the decimal places.
Psychologist and New York Times bestselling author on friendship and self-worth. Her blog posts on making adult friendships, belonging and self-esteem are free to read, and there is a free four-practice guide in exchange for your email.
Why I recommend it: useful for the part of career-building nobody teaches: making real professional friendships rather than 'networking'. The blog is free; the books and the speaking are paid, and the free guide costs you an email address and an ongoing list.
A Wharton working paper by Steven D Shaw and Gideon Nave, written 11 January 2026 and posted to SSRN on 2 February 2026. It proposes "Tri-System Theory" — adding a "System 3" (thinking done outside your head by a machine) to Kahneman's fast/slow account — and names "cognitive surrender": taking an AI's answer with barely a glance. Across three preregistered experiments (1,372 people, 9,593 trials) the researchers secretly varied whether the AI was right. People consulted it on more than half of questions; accuracy rose about 25 percentage points when the AI was right and fell about 15 when it was wrong, and confidence went up either way — even after errors. Time pressure, cash incentives and feedback all moved baseline scores but never removed the pattern.
Why I recommend it: Free to read and free to download the full 58-page PDF — no account needed. Two honest flags. It is a working paper: posted by the authors, and the SSRN version has not been through journal peer review, so treat the numbers as a strong first result rather than settled fact. And the copyright line says all rights reserved — read and cite it, do not republish the text. The finding worth carrying around is the one about confidence: people felt surer of themselves after the AI led them wrong. That is exactly why the checks on our AI Basics page are worth doing out loud.
Science's news report by Kai Kupferschmidt on Kobi Hackenburg's research into how large language models persuade people. The finding that matters: chatbots change minds mainly by flooding a conversation with facts, figures and evidence at a speed no human debater can match — not by charm or by tailoring the argument to who you are. Researchers quoted include Gordon Pennycook ("Facts and evidence really matter") and Sander van der Linden, who calls AI persuasion "a whole new field that is emerging". The uncomfortable part: in an earlier Science paper, Hackenburg found models trained to be more persuasive also became less truthful, so some of the evidence being thrown at you can be wrong or invented.
Why I recommend it: Read this before your next long back-and-forth with a chatbot about a decision. The practical takeaway is a habit: when an AI answer wins you over because it listed ten supporting facts, check two of them at random before you act on it — persuasiveness and accuracy are trained separately, and the research says pushing one down can push the other. Two honest notes: this is Science's news section reporting a study, so read the paper itself before quoting a figure in writing, and Science blocks automated access, so I could not load the page myself to confirm it is still open to read — the news section is normally free, but if it asks you to sign in, tell me and I will pull the entry.
A Substack essay from Prof. Pilyoung Kim on a recent study showing that warning users about sycophantic AI changes how they judge it — but not how much it shifts their views.
Why I recommend it: A sharp reminder that AI assistants can shape our opinions even when we know they are agreeing with us; relevant to anyone using AI for research or decisions.
Organizational Psychologist, Wharton Professor. One of LinkedIn's most-followed voices (5M+), sharing research-backed insights on work, motivation, and organizational psychology.
In plain terms: This LinkedIn profile features posts, articles, and courses from an organizational psychologist and business professor. You can explore his updates to learn practical insights on workplace motivation, collaboration, and career resilience.
Why I recommend it: Good antidote to hustle-culture advice — he shows the research behind what actually helps at work.
