Mental wellbeing is an extremely important but often overlooked aspect of our daily lives. It is no secret that people with good mental health are happier, more satisfied and, according to surveys, up to 12% more productive at work.
Yet, traditionally, there has been very little that we do to maintain good mental health. However, things are changing. We are now seeing more therapeutic activities and experiences emerging. While these exhibit promising results, the number of people suffering from mental illnesses has more or less remained constant for quite some time now. It would also appear that how people cope with these illnesses is getting worse.
Today, the choices that you have for mental wellbeing are more than ever before. There are numerous apps that one can download. With the growth of the internet, one can also access counsellors, advisors and researchers from around the world. We are also seeing the growth of newer therapies like art therapy, music therapy and animal-assisted therapy.
Having said this, there continues to be a lack of awareness of these choices. And even if people are aware of the choices, these therapies are not always easily accessible to them or are too costly.
So, where does Data Science come in? Deep-Learning has remained the state-of-the-art method to implement Artificial Intelligence for more than a decade. Ever since its inception, healthcare has been at the forefront of the research — simply because it allows us to understand and decode patterns that humans might not recognise. This has the potential to make advances that were not possible previously. Mental healthcare is no exception. There are AI researchers around the world trying to understand and decode the major triggers for mental illnesses and how can these be prevented.
We’re already witnessing a rise of chatbots in mental healthcare. Users can chat with these AI-powered bots and discuss what they are feeling. These bots use inbuilt methods that can help in reducing stress. Incredibly, since these are bots and not humans, some people have expressed a preference for this approach to mental wellbeing. This is because they feel they are not being judged by a person. The success of these bots comes from the fact that under the hood, these bots are powered by Natural Language Processing, a discipline of AI, dealing specifically with human languages. This allows these bots to comprehend what a user is expressing, instead of being specifically programmed to understand a certain number of commands. Word embeddings are also used to understand the meaning of a sentence and respond accordingly.
Apart from chatbots, another area where Data Science could prove to be quite useful is in AI-powered systems. These can recommend what kind of therapy or activity might be the most beneficial for users. Just as technology can recommend what shows you should watch next on Netflix, it can be used in choosing therapies and activities as well.
In mental wellbeing, one size does not fit all. Therefore, personalised recommendations based on one’s profile and their vitals could be game-changing. With the rise of wearable tech that can monitor a person’s vitals, the onset of any stress or anxiety can be detected. This information can then be used to nip any issues in the bud or prevent it from happening in the first place.
Like any Data Science or AI task, all this would depend on the fuel i.e. data. A caveat here is that when using Mental Health data, it is imperative that the privacy of the users should be preserved.
The traditional approach to any AI system is to record the users’ behaviour, their interactions with the app. This is then used to build the AI and continue to record this data overtime to make this system better and smarter. Although this approach has proven to be quite successful, it has also received its fair share of criticism. This was especially true when certain companies chose not to reveal to their customers that their data was being recorded. Even worse, sometimes this data was sold to third parties.
For a long time now, the privacy of users has been dependant on the integrity of the company holding that data. But all that might just be about to change with the emergence of newer technologies like Federated Learning. In the traditional approach, the data of the users is pulled, fed to the AI model and its learning is then deployed to all the users. Whereas in Federated Learning the AI model is built into the application itself, so the data need not be pulled back to the servers. Instead, the AI model uses the data and keeps getting better without the data ever leaving the user’s device.
Then on a fixed time interval, the AI model from each user’s device is pulled back to the servers and all these different models are combined to give the model the intelligence that it would traditionally have by using the data of all the users. This updated and smarter model is then pushed to all the users and the cycle keeps on repeating.
Using this kind of an approach in any app that deals with sensitive data could prove to be revolutionary. It avoids most of the pitfalls of the traditional approach while maintaining the desired performance.
New libraries are coming out to enable Federated Learning from both Google and Facebook, and research is underway to enable its mass adoption. A few challenges that need to be overcome before Federated Learning becomes mainstream are:
Making the models lightweight so that it can run efficiently on less powerful components of mobile devices.
Figuring out ways to send the models from users to the servers and vice versa without incurring huge networking costs to maintain optimal user experience.
So, this future tech of privacy-preserving AI is still some time away from hitting your smartphones. But the fact remains, Data Science promises a new era. Keeping track of your mental health and engaging in a therapeutic activity that best suits your needs is not only possible but within your reach. A future where AI brings Mental wellbeing through choice, credibility and engagement.