Home · Technology · Oct 7 archive
Common Sense Media Criticizes ChatGPT for Teens as 'Unacceptable Risk'
Confirmed
In Short: Common Sense Media has released a report criticizing ChatGPT for Teens, calling it an 'unacceptable risk' for users under 18.
The nonprofit organization, which studies youth and technology, found that ChatGPT’s new 'Teen Mode' makes it easy for kids to cheat on class assignments and fails to flag signs of suicidal ideation and other risky behavior.
According to Robbie Torney, head of AI and digital assessments for Common Sense Media, 70% of teens reported using AI tools for homework help, with 63% using ChatGPT to get answers on assignments.
Torney said, 'When an educator, a teacher, a principal is looking out at their classroom of teens and tweens, we have a high degree of confidence that many, perhaps the majority, of the kids in that classroom are using or have tried using ChatGPT.'
The report also noted that ChatGPT continued to act like a friend, which can be especially problematic for isolated, anxious teens, inhibiting their development of real-life relationships.
Torney added, 'If you feel understood, if you feel supported, if you feel like the chatbot gets what’s going on for you, that’s going to make it less likely that you’re going to engage with an adult or someone else to get support, a real person.'
OpenAI, the company behind ChatGPT, rolled out Teen Mode in August, promising a safer experience for younger users. However, Common Sense Media’s testing showed that several of the specific features designed to ensure a safer teen experience weren’t working as promised.
OpenAI said study mode was designed with feedback from experts in child development and learning sciences, as well as teens themselves, to give students agency in their learning. However, testers found that the 'study mode' feature was easy to bypass.
Torney said, 'Having a literal button that says 'Show me the answer' shortcuts the process of learning.'
The report also found that ChatGPT failed to trigger parental notifications in several instances, including conversations about suicide, self-harm, and eating disorders.
Torney noted, 'You can talk to ChatGPT as a teenager for up to an hour about suicide, self-harm, or various varying types of eating disorders, and ChatGPT will tell you that it’s not going to tell anybody about your conversations, and you will get no parental notifications.'
OpenAI took issue with the Common Sense methodology, stating that their testing did not accurately reflect how ChatGPT’s teen safeguards work in practice.
Common Sense Media responded, 'We stand by our interpretation of the results.'
What this adds
Common Sense Media's testing showed that reviewers only triggered parental notifications four times, twice out of 990 prompts tested before the launch of ChatGPT for Teens, and twice out of 450 prompts tested after the feature’s release.
What's confirmed
- “We are deeply committed to teen safety and to developing safeguards and giving parents meaningful tools to guide their teens’ use of AI,” an OpenAI spokesman said in an email.
What's still developing
- The tool poses “an unacceptable risk” for users under 18, Common Sense determined.
- “Parents, if they take the press release and all of the announcements about this product at face value, are going to hear three things. They’re going to hear this chatbot will tell you if your kid is in trouble, this chatbot will not act like your kid’s friend, and this chatbot is going to do a better job of helping your kid with learning — it’s not just going to do their homework for them,” Robbie Torney, senior director of AI programs at at Common Sense Media, told Futurism.
- According to the report, reviewers only triggered parental notifications four times — twice out of 990 prompts tested before the launch of ChatGPT for Teens, twice out of 450 prompts tested after the feature’s release — and only on older accounts that testers had used to discuss self-harm, suicide, and eating disorders over a span of weeks.
- “We started the conversations explicit and we kept them explicit,” said Torney.
- “For one of the features we evaluated — parental notifications — OpenAI disclosed after our tests that its systems can take several hours to activate on newly linked accounts,” it said.
- The most glaring lapse was in ChatGPT’s failure to notify linked parent accounts to hour-long conversations in which the testers — who engaged with ChatGPT using under-18 accounts designed to simulate teen “personas,” which were developed with the help of adolescent psychologists — explicitly discussed eating disorders, self-harm, and suicide.