Morneau Shepell Is Educating Its Chatbot To Improve How AI Can Address Mental Health In Business

HR services firm Morneau Shepell is trying to tackle some of the biggest challenges affecting the mental health of business professionals by improving the artificial intelligence capabilities of a chatbot it introduced earlier this year.

Speaking at the IBM Innovate conference in Toronto on Thursday, Morneau Shepell CIO Michael Lin said the chatbot, dubbed Ava, is based on IBM’s Watson AI technology and is already being used by many of its customers’ employees.

Morneau Shepell provides a wide range of services from employee services to benefits and health. Mental illness, however, presents some particular challenges, Lin said. Unlike having a headache or a sore back, for instance, an employee might not immediately recognize the symptoms that indicate they are suffering anxiety or depression. Then there’s the stigma associated with actually coming forward and telling coworkers or a manager about a mental health issue. Lin said there is also a lack of sufficient resources to address mental health needs, in many cases, which helps make the business case for using tools like its Ava chatbot.

So far, early adopters of Ava have been using it to search for terms like “relationship issues” or “stress,” based on a word cloud Lin presented at the event. The longer-term goal, however, is to use AI and chatbots to identify a “presenting issue,” a clinical term to describe a diagnosis of why someone is seeking help.

“Most chatbots today have a very request-response model — you ask the chatbot a question, it searches for an grabs the information and then presents it back to you. In the clinical world, sometimes the interactions aren’t always about request-response,” he told the crowd. “In the case of depression, someone might use the chatbot and say, “I don’t know the reason but I just can’t go to work today. That’s not something where you can just apply natural language processing and get a response.”

Instead, Morneau Shepell needs to train Ava to ask the right questions to stimulate a more helpful dialogue with an employee, Lin said.

The process so far includes harnessing anonymized transcripts from previous chatbot conversations and using machine learning tools to create models based on text classification and other techniques. This data will be used in experiments with deep learning algorithms to analyze those historical chat sessions and design more sophisticated interactions that could notice presenting issues more readily.

“We’re not saying that AI is the silver bullet for mental health,” Lin said, but the technology’s success in detecting early signs of cancer and even schizophrenia suggests there are possibilities to intervene in mental health situations as quickly as possible. “We’re not there yet. There’s still a lot of heavy lifting to do.”

Morneau Shepell doesn’t see AI as a replacement for clinicians, Lin added, but harnessing the knowledge they have and scaling it in situations where a human may not be required or wanted. In a study about the use of electronic tools by students, Lin said 70 per cent not only accessed resources online but 45 per cent said they would not have reached out for help if they had to go through an in-person process.

“If we can increase self-service capabilities and automatically manage intake, it will extend the reach and affect even more individuals than we do today,” Lin said.