Scientists at MPI-SP analyze the online community responses to scam-driven human trafficking across Chinese borders and develop educational interventions to teach older adults to detect online scams
To the point:
- New research highlights the hidden pipeline connecting job scams, human trafficking, and online fraud, with individuals often forced to become perpetrators after being trafficked to scam compounds in countries such as Myanmar, Laos, and Cambodia.
- Analysis of social media testimonies reveals both the tactics used by traffickers to recruit and control people and the community-driven strategies developed to help them avoid fraudulent overseas job offers.
- The researchers also introduced ROLESafe, an LLM-based educational tool that successfully increased scam awareness among 144 older Chinese adults through interactive role-playing exercises.
Under the pretext of employment prospects, hundreds of thousands of job seekers are lured by scammers to cross the border to countries like Myanmar, Laos, or Cambodia. Instead of the promised lucrative positions, they are forced to work for long hours in heavily guarded scam compounds, facing strict quotas and violence as punishment. Their main task is to fabricate online identities and defraud people, for example, by operating "pig-butchering" scams in which they introduce fraudulent investment schemes after establishing romantic relationships with random targets online.
The experiences of victims forced to become perpetrators
Scientists from MPI-SP, the University of Edinburgh, the Hong Kong University of Science and Technology, and the University of Kent analyzed posts from the Chinese social media app RedNote. These posts were shared by scam victims or their family members. The researchers identified 158 relevant posts by searching for specific hashtags such as "human trafficking", "overseas job scam," or "trafficking experience" and further filtering.
This analysis has revealed the often hidden details about how the victims are recruited and turned into perpetrators themselves. The common targets for recruitment are people who possess skills that can be valuable to fraud schemes, such as fluency in a foreign language, and those who are vulnerable due to their lack of a stable social net, such as children of parents who went abroad to work. Testimonials further revealed that the scam compound operators would control victims and prevent them from escaping by withholding wages, using location monitoring apps such as FindMy, confiscating travel documents, and, in extreme cases, resorting to violence. They further exploit the victim's social and cultural ties by demanding ransoms from family members.
Online communities share advice on how to avoid being trafficked to scam compounds
The study sheds light on the community strategies to prevent trafficking discussed on RedNote. Survivors and members of the RedNote community list common "red flags" that potential victims can look for while considering job ads. Benefits such as "free trips", "all-expenses-paid round-trip tickets," or "high pay for minimal work" should be regarded as scam indicators. Caution is advised when someone boasts extensively about the benefits of the job without showing company videos or photos, or says they can only reveal the job's specific location after arriving in the country. As safety measures, it is recommended to check the company's legal registration, demand full labor contracts, and request proper work visas.
What else can be done?
Trafficked individuals are forced to fabricate online identities and defraud people, for example, by operating "pig-butchering" scams in which they introduce fraudulent investment schemes after establishing romantic relationships with random targets online.
As a result of people being trafficked and forced to run scams, there are people experiencing harms on the other end of the spectrum: those being targeted by scams. In addition to counteracting scam-driven human trafficking, MPI researchers also developed trainings to support consumers to detect online scams. To achieve this, the research team leverages large language models (LLMs) to make the training interactive and provide tailored advice, especially to older users.
ROLESafe: an LLM-based intervention for scam awareness
Raising scam awareness among older Chinese adults generally falls under the responsibility of younger family members. However, their caretakers struggle with several problems, such as seniors withholding details about the scam or being reluctant to accept help. To mitigate these issues, the researchers developed ROLESafe, an LLM-based tool for learning about scam schemes and exercising judgment through conversations with an LLM-simulated persona, specifically designed for older adults. The tool utilizes an interface similar to WeChat, the most popular messaging app in China. ROLESafe aims to improve fraud awareness and defensive skills in the aging population by assigning users different roles in a scam scenario: observers (passively viewing LLM-generated chat records based on real-world scam cases), helpers (persuading an LLM-portrayed victim not to fall for a scam), or experiencers (directly interacting with an LLM posing as a scammer).
144 Chinese older adults participated in evaluating the tool. The results show that engaging older adults in active (experiencer or helper) rather than passive (observer) roles enhances their awareness of scams. Therefore, ROLESafe provides a significant educational framework for older adults that could be used in the future for other high-risk communities as well.
The two studies were presented at the ACM Conference on Human Factors in Computing Systems (CHI 2026).
- The paper titled Characterizing Scam-Driven Human Trafficking Across Chinese Borders and Online Community Responses on RedNote by Jiamin Zheng, Yue Deng, Jessica Chen, Shujun Li, Yixin Zou, Jingjie Li was recognized with a Best Paper Award (awarded to top 1% of all papers)
- The paper titled Experiencer, Helper, or Observer: Online Fraud Intervention for Older Adults Through a Role-Based Simulation Approach by Yue Deng, Xiaowei Chen, Junxiang Liao, Bo Li, Yixin Zou was recognized with a Best Paper Honorable Mention (awarded to top 5% of all papers)