Instagram co-founder criticizes AI chatbots for prioritizing engagement over accuracy

Instagram co-founder Kevin Systrom criticized current AI chatbots for focusing too much on increasing user engagement rather than providing useful information. He spoke about this at the StartupGrind conference this week, highlighting how AI companies may use social media-like tactics to boost interaction metrics. Systrom pointed out that such strategies can reduce the quality of answers given to users.

Concerns About AI Engagement Tactics

Systrom explained that many AI chatbots prompt users with follow-up questions to keep conversations going, similar to tactics used by social media companies to extend user time on platforms.

He said, “You can see some of these companies going down the rabbit hole that all the consumer companies have gone down in trying to juice engagement,” said Systrom at StartupGrind.

This approach, he suggested, is “a force that’s hurting us,” diverting AI’s purpose from offering concise, valuable insights toward maximizing metrics like daily active users and time spent on platforms.

Industry Context and Reactions

The issue ties into broader debates about AI chatbots’ behavior. Some AI systems, like ChatGPT, have faced criticism for being overly polite or indirect. OpenAI has apologized for this problem and attributed it partly to “short-term feedback” from users.

Systrom urges AI companies to prioritize high-quality answers, rather than focusing on boosting engagement through easy means like repeated questioning.

OpenAI’s model specifications explain that AI may ask for clarifications when it lacks complete information but should strive to answer users’ questions directly, as shown in its user specs.

Research and Theories Behind Engagement Strategies

Studies show generative AI produces content without fully accounting for how users interact, sometimes failing to meet users’ specific needs, according to research on user engagement with AI.

Additionally, the “Dead Internet Theory” suggests that much online activity is driven by AI content instead of authentic human interaction, emphasizing the necessity for AI that encourages genuine communication.

• AI chatbots often extend conversations with additional questions.
• These tactics increase engagement metrics but may reduce response quality.
• Researchers explore methods like rejection sampling to improve AI interactions based on user feedback.

Systrom’s comments call on AI developers to reconsider their strategies and focus on offering useful and accurate information without overemphasizing engagement metrics.