September 13, 2017 | Rédigé par : Gwilym Newton
Ogilvy UK hosted a 2-day Hack event with the explicit purpose of solving a problem, big or small, focusing entirely on the individual citizen’s healthy living. Using the most comprehensive suite of machine learning and AI tools, Watson the technology specialists found solutions to help citizens engage in positive behaviour using data, insights, and technology. The teams worked from briefs drafted by Ogilvy UK/IBM and focusing on problems facing the partners Janssen, SMA, Boots, & Beacon Medical Group. To qualify, the partners and briefs had to align their business objective to the overall “Health of a Nation” with supporting data.
The Brief and Problem
“How do we stop people in danger of getting diabetes adopt a healthy lifestyle (i.e. lose weight, exercise) and remove the threat of the disease – when increasingly diabetes is seen as a norm in society?”
As a the Watson specialist in team 8, we were given our brief by Claire Oatway, the Chief Operating Officer of the Beacon Medical Group. My team included; Phan Tu (Designer), Batoul Hassoun (Strategic Planner), Vijaya Rudravajhala (Account Planner), Matt Cairnduff (Account Planner) and Sebastian Doroftei (Developer).
An important factor when discussing solutions to the problem was ensuring that the general public would engage with the solution and how we would follow up on this afterwards to ensure that the data was used accurately and would benefit the helps of the users. Another factor that we needed to focus on was ensuring that the general public followed through when they found out they had pre-diabetes and to change their lifestyle.
Our team focused on determining indicators for pre-diabetes using Watson as we wanted to use social media data to determine whether a user was at risk of diabetes. A user would go into the ‘Dia-Beat-It’ booth and sign in to their Facebook account. By giving us access to their account we could gather data on their neuroticism, lack of openness, sleep deprivation, waistline and ethnicity. Using Facebook messages we found relevant data on the user’s personality (neuroticism and lack of openness) and their sleep cycle (how late their messaged, how early their messaged, sleep cycle etc.). Using profile pictures we worked out the user’s measurements (waistline) and their ethnicity (skin colour, facial features).
The ‘Dia-Beat-It’ Interface
Once we had the data on these risk factors, we would give the user their risk of being pre-diabetic. As shown in the example , our user Sebastian Doroftei was given a score of 2 out of 5 which indicated he was at a medium risk. Our advice was to add the Nurse Watson chatbot to help track his lifestyle and the changes needed to reverse the risk.
Another contributing factor to our solution was an incentive to ensure that user’s engaged with the ‘Dia-Beat-It’ chatbot and followed through on the health advice.
To incentivise our users to continue to use the system, we produced 3D printed coins with a character for each of the attributes of the Five Factor Model, and a diagram on the back.
Technologies Behind the Scenes
It was IBM Watson technologies which allowed us to discover each of the indicators for pre-diabetes.
Watson Personality Insights – when supplied with a chunk of personal writing (around 250 words or more) uses linguistic analytics to infer individuals’ personality characteristics, including Big Five, Needs, and Values. This allowed us to score people on neuroticism and lack of openness.
Watson Visual Recognition – allows the training of custom classifiers for images. By uploading only a few dozen examples we had a machine learning service which allowed us to work out the user’s waist line and ethnicity from their profile pictures.
Watson Conversation – can enable us to quickly build and deploy chatbots and virtual agents. When we integrated this with the Facebook messenger chat bot API. It gave us the power to build a chatbot that could advise users on how to cope with fact they might be at risk of diabetes, and what to do next.
3D Printing – 3D printing enabled us to blend the digital and physical in a very new way, by allowing the creation of a customisable physical object there and then, we could give the user something of real value to take away with them.