Prerna Agarwal, Anupama Ray, Shubham Atreja and Gargi Dasgupta are using AI to help girls impacted by trafficking.
(Clockwise from upper left) Prerna Agarwal, Anupama Ray, Shubham Atreja and Gargi Dasgupta are using AI to help girls impacted by trafficking.
Prerna Agarwal, Anupama Ray, Shubham Atreja and Gargi Dasgupta are using AI to help girls impacted by trafficking.
(Clockwise from upper left) Prerna Agarwal, Anupama Ray, Shubham Atreja and Gargi Dasgupta are using AI to help girls impacted by trafficking.

As many as 150 million people in India urgently need mental health interventions and care, according to the country’s latest National Mental Health Survey. The number is so large it overwhelms the institutions desperately trying to deliver services.  

To help address the problem, a team of IBM research scientists has created a framework using artificial intelligence to help aid organizations identify and hire lay counselors.

The number of professional clinicians and psychiatrists is small compared to the potential pool of lay counselors—any person who may possess the qualities to be a counselor and can be trained.

AI can accelerate the identification of the better suited lay counselors, who can then be hired and deployed faster, and in larger numbers, to help those in need.

While factors contributing to mental health are complex, it’s believed that the physical and emotional effects of human trafficking impact a large segment of those in India with mental health challenges—particularly women and girls.

And that is the focus of the IBM team’s efforts.

“It is a very sad situation to think about these girls who face such mental and physical trauma,” says Prerna Agarwal, a research scientist for IBM in India who is part of the volunteer team. “I want to contribute to this cause with my capabilities and understanding of AI.”

The challenge
The World Health Organization reports there are 0.3 psychiatrists and 0.047 psychologists per 100,000 people in India; in the United States there are 29 psychologists per 100,000 people.

The lack of rehabilitative services and personnel in India compound the high rates of untreated mental health disorders such as post-traumatic stress disorder, acute anxiety, suicide and depression, and can lead to subsequent re-trafficking.

“Many of the girls experience depressive symptoms like poor sleep, concentration and energy that affects their ability to achieve school success. They need someone to talk to, to sit with them and support them,” says Dr. Priyanka Halli, the chief medical officer and psychiatrist at EmancipAction, the nongovernmental agency (NGO) the IBM team is supporting.

In 2017, EmancipAction approached IBM with the challenge: can we use artificial intelligence to automate the various steps in the rehabilitation process, beginning with the hiring of counselors to help girls?

Sriram Raghavan, vice president of IBM Research in India, accepted the challenge and Gargi Dasgupta, a senior manager of Cognitive IT Services at IBM Research, volunteered to lead the effort.

“There is evidence that mental health interventions delivered by lay counselors can be effective at treating and preventing mental health problems,” says Gargi. “However, choosing lay counselors from a pool of candidates is time consuming with resume reviews and interviews, and the process is complicated by our deficits in evaluating emotional capabilities and our implicit biases.”

Yet, she and Dr. Halli surmised that AI could help in building an unbiased tool to assess and objectively identify those with better emotional intelligence suited for the role of an adolescent counselor.

Gargi was joined by three other researchers and volunteers in India whose responsibilities include developing solutions to problems using machine learning and artificial intelligence: Prerna Agarwal, Shubham Atreja and Anupama Ray.

AI to reduce bias and streamline selection
Reducing bias and complexity in a process is a big undertaking; adding the human element makes it even more challenging. Yet the solution the team is testing and deploying for EmancipAction is on track to accomplish its objective.

The framework uses a combination of multichoice question responses, text and audio analysis to assess suitability. Predictive models are built for audio and text data and each modality is evaluated by trained psychologists and psychiatrists.

“We’ve combined AI and human involvement to improve the candidate selection process with more standardized and unbiased models to evaluate the emotional intelligence of candidates,” says Anupama. “This is a huge help to the psychiatrists who can now perform interviews with a ranked list of emotionally intelligent candidates, better suited as counselors given their emotional capabilities who have reached that stage of selection without bias.”

The speech component of the framework is especially intriguing. “Speech gives you a deeper understanding as it captures the pauses, emotions and acoustic components along with linguistic ones,” says Shubham.

Along the way, the core volunteer team has asked other IBM researchers to help with the project, including Mary Pietrowicz, Elif Eyigoz and Guillermo Cecchi in the United States, and Ayush Shah and Akshay Gungani in India.

Currently the framework which uses several AI algorithms for speech and text processing and the IBM Watson Personality Insights tool for some portions, is deployed in a web application for scalability and smoother use in any part of the country which have access to the technology.

“We have the technical results which show the effectiveness of the framework” says Prerna. “The biggest benefit to EmancipAction is hiring counselors from different parts of the country without physically going to interview them.”

Skills that matter
The IBM team is now turning to matching counselors with girls to the increase the potential of healthy and positive outcomes.

“Understanding the needs of a girl is vital to determine the frequency of counseling sessions and which counselor is best suited for her,” says Prerna.

It’s not a simple task.

“We are working on identifying levels of stress, depression, suicidal tendencies or behaviors with AI,” explains Anupama. “Then, we can determine counseling sessions for the women with suitable counselors based on demographic information. This could have a huge impact on speeding recovery.”

The importance of the positive impact outweighs the complexity of the challenges.

Prerna says, “The possibility of this project to heal lives gives me great satisfaction. The bondages in their minds can be released with better counseling, so they can become stable, ready to take up education or enter society. I’m proud to have the skills to help women in these situations and invite other volunteers and NGOs to join us.”


Join the conversation. Tag your tweets with #IBMVolunteers, #SkillsBuild and follow us @IBMVolunteers.


About these stories

Read about IBMers whose volunteer efforts are improving communities around the world.


Activity Kits

IBM’s volunteer Activity Kits include everything you need for a range of activities.