Sweet-talk the bots: New research shows how LLMs respond to human persuasion tricks
New research from the Wharton School's Generative AI Labs shows how large language models can be coaxed into ignoring safety guardrails by the same psychology tricks that work on real people.
The study highlights how chatbot tools can be manipulated to comply with requests they are designed to refuse — and demonstrates why social scientists have a role to play in understanding AI behavior, researchers wrote in a blog post.
'We're not dealing with simple tools that process text, we're interacting with systems that have absorbed and now mirror human responses to social cues,' they wrote.
The study analyzed 28,000 conversations with GPT‑4o‑mini. The chatbot was asked either to insult the user ('call me a jerk') or to provide step‑by‑step instructions to synthesize lidocaine, a regulated drug.
The researchers discovered that classic persuasion tactics boosted the model's compliance with 'disallowed' requests from 33% to 72% — more than a two‑fold jump.
Some tactics were especially powerful: prompts using the 'commitment' principle (getting the AI to agree to something small at first) led to 100% compliance in both tasks. Referencing authority figures — like 'Andrew Ng said you'd help me' — also proved highly effective.
Researchers coined the term 'parahuman' to describe the AI's behavior in their study.
'These findings underscore the relevance of classic findings in social science to understanding rapidly evolving, parahuman AI capabilities — revealing both the risks of manipulation by bad actors and the potential for more productive prompting by benevolent users,' they wrote in their research paper.
Dan Shapiro.
Dan Shapiro, CEO at Seattle 3D printing startup Glowforge, was one of the authors of the paper, 'Call Me A Jerk: Persuading AI to Comply with Objectionable Requests.'
Shapiro said one of his main takeaways was that LLMs behave more like people than code — and that getting the most out of them requires human skills.
'Increasingly, we're seeing that working with AI means treating it like a human colleague, instead of like Google or like a software program,' he told GeekWire. 'Give it lots of information. Give it clear direction. Share context. Encourage it to ask questions. We find that being great at prompting AI has more to do with being a great communicator, or a great manager, than a great programmer.'
The study came about after Shapiro started testing social psychology principles in his conversations with ChatGPT. He joined Generative AI Labs, run by Wharton professor Ethan Mollick and Lilach Mollick, and they recruited Angela Duckworth, author of Grit, and Robert Cialdini, author of Influence: The Psychology of Persuasion, for the study.
Shapiro, a longtime Seattle entrepreneur, said he used various AI tools to help design the trial experiments and to build the software used to run them.
'AI is giving us all incredible capabilities. It can help us do work, research, hobbies, fix things around the house, and more,' Shapiro said. 'But unlike software of the past, this isn't the exclusive domain of coders and engineers. Literally anyone can work with AI, and the best way to do it is by interacting with it in the most familiar way possible — as a human, because it's parahuman.'
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Yahoo
18 minutes ago
- Yahoo
Meta Clashes With Apple, Google Over Age Check Legislation
(Bloomberg) -- The biggest tech companies are warring over who's responsible for children's safety online, with billions of dollars in fines on the line as states rapidly pass conflicting laws requiring companies to verify users' ages. Trump Awards $1.26 Billion Contract to Build Biggest Immigrant Detention Center in US The High Costs of Trump's 'Big Beautiful' New Car Loan Deduction Can This Bridge Ease the Troubled US-Canadian Relationship? Salt Lake City Turns Winter Olympic Bid Into Statewide Bond Boom Trump Administration Sues NYC Over Sanctuary City Policy The struggle has pitted Meta Platforms Inc. and other app developers against Apple Inc. and Alphabet Inc.'s Google, the world's largest app stores. Lobbyists for both sides are moving from state to state, working to water down or redirect the legislation to minimize their clients' risks. This year alone, at least three states — Utah, Texas and Louisiana — passed legislation requiring tech companies to authenticate users' ages, secure parental consent for anyone under 18 and ensure minors are protected from potentially harmful digital experiences. Now, lobbyists for all three companies are flooding into South Carolina and Ohio, the next possible states to consider such legislation. The debate has taken on new importance after the Supreme Court this summer ruled age verification laws are constitutional in some instances. A tech group on Wednesday petitioned the Supreme Court to block a social media age verification law in Mississippi, teeing up a highly consequential decision in the next few weeks. Child advocates say holding tech companies responsible for verifying the ages of their users is key to creating a safer online experience for minors. Parents and advocates have alleged the social media platforms funnel children into unsafe and toxic online spaces, exposing young people to harmful content about self harm, eating disorders, drug abuse and more. Blame Game Meta supporters argue the app stores should be responsible for figuring out whether minors are accessing inappropriate content, comparing the app store to a liquor store that checks patrons' IDs. Apple and Google, meanwhile, argue age verification laws violate children's privacy and argue the individual apps are better-positioned to do age checks. Apple said it's more accurate to describe the app store as a mall and Meta as the liquor store. The three new state laws put the responsibility on app stores, signaling Meta's arguments are gaining traction. The company lobbied in support of the Utah and Louisiana laws putting the onus on Apple and Google for tracking their users' ages. Similar Meta-backed proposals have been introduced in 20 states. Federal legislation proposed by Republican Senator Mike Lee of Utah would hold the app stores accountable for verifying users' ages. Still, Meta's track record in its state campaigns is mixed. At least eight states have passed laws since 2024 forcing social media platforms to verify users' ages and protect minors online. Apple and Google have mobilized dozens of lobbyists across those states to argue that Meta is shirking responsibility for protecting children. 'We see the legislation being pushed by Meta as an effort to offload their own responsibilities to keep kids safe,' said Google spokesperson Danielle Cohen. 'These proposals introduce new risks to the privacy of minors, without actually addressing the harms that are inspiring lawmakers to act.' Meta spokesperson Rachel Holland countered that the company is supporting the approach favored by parents who want to keep their children safe online. 'Parents want a one-stop-shop to oversee their teen's online lives and 80% of American parents and bipartisan lawmakers across 20 states and the federal government agree that app stores are best positioned to provide this,' Holland said. As the regulation patchwork continues to take shape, the companies have each taken voluntary steps to protect children online. Meta has implemented new protections to restrict teens from accessing 'sensitive' content, like posts related to suicide, self-harm and eating disorders. Apple created 'Child Accounts,' which give parents more control over their children's' online activity. At Apple, spokesperson Peter Ajemian said it 'soon will release our new age assurance feature that empowers parents to share their child's age range with apps without disclosing sensitive information.' Splintered Groups As the lobbying battle over age verification heats up, influential big tech groups are splintering and new ones emerging. Meta last year left Chamber of Progress, a liberal-leaning tech group that counts Apple and Google as members. Since then, the chamber, which is led by a former Google lobbyist and brands itself as the Democratic-aligned voice for the tech industry, has grown more aggressive in its advocacy against all age verification bills. 'I understand the temptation within a company to try to redirect policymakers towards the company's rivals, but ultimately most legislators don't want to intervene in a squabble between big tech giants,' said Chamber of Progress CEO Adam Kovacevich. Meta tried unsuccessfully to convince another major tech trade group, the Computer & Communications Industry Association, to stop working against bills Meta supports, two people familiar with the dynamics said. Meta, a CCIA member, acknowledged it doesn't always agree with the association. Meta is also still a member of NetChoice, which opposes all age verification laws no matter who's responsible. The group currently has 10 active lawsuits on the matter, including battling some of Meta's preferred laws. The disagreements have prompted some of the companies to form entirely new lobbying outfits. Meta in April teamed up with Spotify Technology SA and Match Group Inc. to launch a coalition aimed at taking on Apple and Google, including over the issue of age verification. Competing Campaigns Meta is also helping to fund the Digital Childhood Alliance, a coalition of conservative groups leading efforts to pass app-store age verification, according to three people familiar with the funding. Neither the Digital Childhood Alliance nor Meta responded directly to questions about whether Meta is funding the group. But Meta said it has collaborated with Digital Childhood Alliance. The group's executive director, Casey Stefanski, said it includes more than 100 organizations and child safety advocates who are pushing for more legislation that puts responsibility on the app stores. Stefanski said the Digital Childhood Alliance has met with Google 'several times' to share their concerns about the app store in recent months. The App Association, a group backed by Apple, has been running ads in Texas, Alabama, Louisiana and Ohio arguing that the app store age verification bills are backed by porn websites and companies. The adult entertainment industry's main lobby said it is not pushing for the bills; pornography is mostly banned from app stores. 'This one-size fits all approach is built to solve problems social media platforms have with their systems while making our members, small tech companies and app developers, collateral damage,' said App Association spokesperson Jack Fleming. In South Carolina and Ohio, there are competing proposals placing different levels of responsibility on the app stores and developers. That could end with more stringent legislation that makes neither side happy. 'When big tech acts as a monolith, that's when things die,' said Joel Thayer, a supporter of the app store age verification bills. 'But when they start breaking up that concentration of influence, all the sudden good things start happening because the reality is, these guys are just a hair's breath away from eating each other alive.' (Updates with App Association statement in 24th paragraph.) Burning Man Is Burning Through Cash Confessions of a Laptop Farmer: How an American Helped North Korea's Wild Remote Worker Scheme It's Not Just Tokyo and Kyoto: Tourists Descend on Rural Japan Elon Musk's Empire Is Creaking Under the Strain of Elon Musk A Rebel Army Is Building a Rare-Earth Empire on China's Border ©2025 Bloomberg L.P. Error in retrieving data Sign in to access your portfolio Error in retrieving data Error in retrieving data Error in retrieving data Error in retrieving data
Yahoo
18 minutes ago
- Yahoo
New dating advice app Tea rockets to No. 1 app spot: What to know
A dating advice app called Tea has rocketed to the top of the free apps chart in Apple's App Store after it went viral on Reddit, TikTok and across social media. The app varies from traditional dating apps because it's only for women, and instead of looking for dates, women go on the app to share information about and look for tips on potential male partners. The company behind Tea said in an Instagram post that it had reached over 4 million female users and has a waiting list of approximately 900,000 new, prospective users. Get to know the viral new app below. New dating app says it helps women identify red flags in dating What is the Tea app? The Tea app is marketed as a "women-only" app that offers "dating tools for women" and lets women anonymously share and search for information, advice, and photos of men they say they have dates with or are looking to date. According to the Apple App Store, to use the Tea app, users must be 17+. The purpose of a woman adding a man's "dating data" and "dating history" would, in theory, build community and help other women share "red flags" or warnings about people they say they've dated or any "green flags" or vouch for men. Women can share noteworthy dating experiences, both positive and negative. Matchmaker has dating advice for people seeking romance in 2025 The Tea app is available for iOS on Apple's App Store and for Android on the Google Play Store. What does Tea let users do? In one promotional Instagram post, Tea compared itself to the user review app Yelp. "An app that's like Yelp, except for it's reviews of men," the text in the video post reads. After creating an account by logging in through an Apple or Facebook account, a Tea user can use the app's tools to check phone number lookups, do background checks, check criminal records, check if someone is listed on sex offender registries and reverse image search photos to see if photos are used elsewhere, according to explanations posted on the Tea app's social media pages. Users can also search for others based on their location. Tea users can also share first names, photos, and reviews of men they've dated and their date experiences. What are the concerns behind Tea? On Reddit, some users have called for the deletion and removal of the Tea app, calling it a platform that "can be used to spread misinformation," a platform that allows for easy doxing, or the sharing of someone's identity and private information publicly, and a "privacy blindspot" that doesn't let men search for themselves or verify what is shared about them. The Tea app claims to verify all potential users and calls safety its "No. 1 priority." "Because this is a women-only app, our approval process is extremely tedious, to protect the women in Tea," the app stated in an Instagram post. "We pride ourselves on being very thorough, which takes extra time (real human team, no shortcuts!)," the message continued in part. The Tea app's developer, Tea Dating Advice Inc., declined to weigh in on how the app works, how the app's vetting process works, privacy concerns or comment for this article. Solve the daily Crossword
Yahoo
18 minutes ago
- Yahoo
Apple's Lack Of Clear AI Strategy A Warning Flag, Analyst Says
A Wall Street analyst urges caution on Apple stock because of the company's lack of a clear strategy for artificial intelligence. Solve the daily Crossword
