AI for Mental Healthcare
A collaboration between therapist, client, and Artificial Intelligence for NiceDay
The applications of AI are multifold across almost every industry. In the mental healthcare sphere, in particular, AI has the potential to make care more personalized and need-driven. However, in order to enable the adoption of AI in mental healthcare, it is vital for decision-making AI systems to be understandable for both developers, end users, and the people that are subjected to an AI-enabled decision.
Deeploy has supported NiceDay, a leading (online) mental health care platform, in this journey, helping keep their AI systems transparent, explainable, and trustworthy. Mental health therapists are supported by AI in their daily decision process, helping them decide who to reach out to and at which moment in time, based on the client’s personal needs.
NiceDay is a leading (online) mental healthcare platform in the Netherlands, helping therapists treat clients more effectively and ultimately strengthening the mental wellbeing of as many people as possible.
NiceDay aims to change the way mental healthcare is provided, using current developments in AI to deliver personal and effective support from therapists in the recovery of people with mental health complaints.
Use case: Supporting therapists with AI to provide need-driven care
The NiceDay platform connects therapist and patient, aiming to foster collaboration between the two. In between sessions, clients can register information about their well-being, upload finished therapy assignments, and write journal entries.
Informed by these registrations, therapists may then decide to check in with a client directly via the chat function in NiceDay. This in-between session forms an opportunity to provide need-driven, personalized mental health care. However, for therapists to decide which clients to reach out to at what moment in time is a challenge, given their limited time and the large amount of information to process.
In order to overcome this, NiceDay has an AI system in place that analyzes the data imputed by therapists and then provides summarized information for therapists to determine when and to which clients to reach out to.
When clients register information, data is collected to be fed into the machine learning model which then presents the therapists with a ranking of which clients to reach out to first.
Nevertheless, while this can ensure that the care becomes more need-driven and personalized, AI adoption in mental heatlcare raises some concerns.
Challenges for using AI within mental healthcare
Privacy & Security
Given the sensitive nature of a patient’s medical data, it is vital to ensure its safety, security, and confidentiality. As such, client data must be anonymized and confidential during the developmental phase of the AI systems and stored with care after deployment.
Cooperation between the AI model and therapists
Moreover, there is also the challenge of making sure the cooperation between AI model and therapists works. The therapist, as the end-user, needs to be able to interact with the model through a simple and intuitive interface.
Besides this, it is also important for therapists to understand the model decisions on a deeper level. This is where explainability comes in. Explainability encompasses a range of different aspects, but, in this context, it refers to the explainability of the decisions made by the model.
Which general trends are represented by the model? Which feature contributed most to the prediction? What input would lead to a different decision?
As such, each recommendation the therapist gets is accompanied by an explainer that gives the full picture of what led to the model’s ranking.
Therapists being able to reach out to clients exactly when they need to is promising. Nevertheless, this won’t be useful unless predictions are accurate. But how can accurate predictions be ensured? By keeping a human-in-the-loop (in this case the therapist) who can partake in the feedback loop.
After getting the ranking recommendation, the therapist can either accept it or overrule it and give feedback on this overruling.
This feedback is then collected and used to improve the model, helping it make more accurate predictions.
The feedback loop continues, allowing the model to become more and more accurate over time and, in turn, streamlining the work of the therapist more and more.
Interested in a more in-depth view of this process? Deeploy and NiceDay developed a white paper document discussing the in depth development of an AI system for NiceDay, using the BAIT method that combines expert input with observational-level choice data.
Feasible usage of AI in mental healthcare
Implementing AI systems in healthcare is challenging. It is critical for the end users to understand and be able to interact with the interface and, most importantly, model decisions must be made explainable and understandable. Armed with enough knowledge about those decisions, end users can then decide to follow the prediction or overrule it and give feedback, which will only help make models more accurate over time.
NiceDay’s AI system was developed in a clear and intuitive manner and with the therapists in mind. With the assistance of AI, therapists can make their processes more efficient and deliver more care driven and personalized assistance to their clients.
Deeploy’s responsible AI platform allows for deploying AI systems in mental healthcare in an easy, transparent and explainable manner.
Want to explore the full potential of Deeploy?