The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence) Human-Centered ML for High-Risk Behaviors with Stevie Chancellor - #472
Apr 5, 2021
Stevie Chancellor, an Assistant Professor at the University of Minnesota, tackles the intersection of human-centered machine learning and high-risk behaviors. She shares insights on using machine learning to assess mental illness severity and discusses how convolutional graph neural networks can reveal new behaviors in opioid use disorder. Chancellor also delves into the ethical challenges of mining social media data for mental health research, underscores the importance of clear communication in mental health, and emphasizes ethical considerations in AI-driven crisis detection.
AI Snips
Chapters
Transcript
Episode notes
Facial Recognition Example
- Facial recognition at airport checkpoints highlights the broad impact of technology.
- Chancellor points out that even those not directly using the system are affected, raising privacy concerns.
Nuance in Online Crisis
- Not all expressions of mental distress online indicate imminent danger.
- Chancellor suggests considering the context and providing appropriate support instead of assuming a need for drastic intervention.
Opioid Use Disorder Research
- Chancellor is using graph convolutional neural networks to study opioid use disorder.
- The goal is to identify new behaviors and alternative therapies discussed in online communities.