Photo: CC Pixabay.
Reading Time: 2 minutes

Research by U of O prof to produce new detection algorithms

Social media could soon be put to good use when it comes to detecting possible mental illness, according to research by University of Ottawa Engineering and Computer Science professor Diana Inkpen and her team.

The project will use data collected from social media websites to produce an algorithm that detects signs linked to possible mental illness.

Inkpen, a computer scientist and engineer, is leading a team of researchers at the U of O in collaboration with the University of Albert and Université de Montpellier in France. They will also collaborate with the company Advanced Symbolics, which offers “predictive analytics,” to collect and sample the data from social media.

While the project is still in its very early stages, Inkpen said some future applications could be to help doctors increase quality of the care provided to patients, detect cyberbullying, and for use by mental illness support services, such as help telephone lines.

“They can train some models on their data and when new calls come in, the model can already offer something learned from the past from other users in similar situations, so more like a research tool or a help tool for the operator or a help line,” said Inkpen.

The team began their research in October 2015, after receiving the notice of the three-year grant from the Natural Sciences and Engineering Research Council of Canada, and have begun to analyze Twitter data, which is public and easily to collect.

The project will use data mining algorithms, where the team will collect data from social media platforms, manually annotate it and then use this data to train algorithms. These algorithms are able to pick up patterns in new data and associate those patterns with possible signs of mental illness, said Inkpen.

“This is the idea of data mining: we learn from, we train the algorithms on training data, save the prediction models and then we can use the models on new data to make predictions in this userface,” said Inkpen.  The main question, she said, “is the user at risk of something?”

In addition to working with other researchers, Inkpen said they are consulting with Dr. William Gardner of the Children’s Hospital of Eastern Ontario (CHEO), who specializes in pediatric mental health.

Gardner said the prospect of the possible benefits of this research “could be extraordinarily valuable,” but cautions not to overestimate the future use of the research within medical practice.

He said while this will not be a diagnostic tool itself, it may be useful in the future to identify those who may need to see a psychiatrist or psychologist.

“You really need to have somebody diagnosed, actually diagnosed by somebody qualified to do so in order to say this person is or is not mentally ill,” said Gardner.

Gardner also raises concerns on the ethics of using this technology in medicine, cautioning that monitoring patients’ social media without their knowledge may be unethical, and questioning if getting their consent to do so would even be possible.

“I think there’s both technical and medical ethical questions that would have to be very carefully explored before we would really be able to say what are the possible applications,” he said.