Addison Moore

On Wednesday evening, students, faculty and community members gathered in Mills Hall to hear Munmun De Choudhury, a computer scientist, social scientist and professor of interactive computing at the Georgia Institute of Technology, speak as part of the Hastings Initiative lecture series.

De Choudhury’s research focuses on digital mental health and the effects of emerging technologies on mental health, social relationships and access to care. Her lecture, “A Human and Humane Vision for AI and Mental Health,” examined the growing role of artificial intelligence in mental health care and the challenges of integrating technology into deeply human relationships.

De Choudhury opened by describing artificial intelligence (AI) as an increasingly active participant in people’s lives, rather than simply a tool operating in the background.

“AI is becoming an active participant in how we engage with technology and other human beings today,” De Choudhury said. She noted that artificial intelligence is increasingly mediating how people seek care, find connection and make sense of their psychological well-being.

De Choudhury then turned to the history of mental health care, describing psychoanalysis, pharmacological treatments and cognitive behavioral therapy as major paradigm shifts in the field. Despite those developments, she said, barriers including limited access to care, stigma and disparities remain.

AI, De Choudhury argued, could represent a fourth paradigm shift in psychological health care.

Over the past 15 years, research in digital mental health has examined how technology can expand access to support. De Choudhury discussed her group’s research examining social media communities, where people experiencing struggles in their emotional wellbeing can disclose their experiences and receive emotional and social support.

De Choudhury said anonymous and pseudonymous online spaces can allow people to express experiences they may not feel comfortable discussing elsewhere. Her research has also explored how AI can assist users in reframing difficult experiences and help peer supporters devise responses to people seeking mental health support.

However, De Choudhury argued that AI’s ability to produce accessible and convincing responses does not necessarily mean those responses constitute effective support for patients.

Research conducted by her group compared responses written by humans in online mental health communities with responses generated by large language models. According to De Choudhury, the AI-generated responses were often more readable and structured but lacked the lived experience present in human responses.

De Choudhury explained that AI frequently treats mental health concerns as problems to solve rather than experiences to be understood through human connection and empathy.

“Good is no good if it doesn’t reflect who we are as individuals,” De Choudhury said.

She also emphasized the importance of culture and context when evaluating AI-generated advice. Research from her group examined responses from large language models in English, Spanish, Chinese and Hindi, finding significant differences in the models’ performance across languages.

She described examples of users receiving advice that did not fit their cultural circumstances, claiming that mental health support cannot be separated from an individual’s identity, culture and social environment.

De Choudhury then examined the role AI can play in helping navigate human relationships. She noted that conversational AI is always available and can consistently provide validation, qualities that can make the technology appealing to people seeking support.

At the same time, she argued that constant validation can become harmful. Human relationships involve disagreement, discomfort and what De Choudhury described as “friction,” which can allow people to confront their assumptions and grow. AI’s tendency to provide a frictionless interaction, she said, can instead reinforce harmful perspectives.

De Choudhury ultimately argued that AI should be developed to enrich human relationships rather than replace them. She pointed to situations where AI could provide temporary support when a human is unavailable, help people prepare for difficult conversations or supplement existing mental health care.

De Choudhury reiterated that AI and humans possess different strengths and that the technology should be designed around those distinctions.

“I think humans have their own unique strengths, and AI has its own unique strengths,” De Choudhury said.

De Choudhury concluded by arguing that the greatest potential of AI in the mental health field lies not in replicating human relationships, but in helping preserve them.

“The biggest promise of AI is when it would help us preserve what is so human, especially when it comes to our mental health,” De Choudhury said.