If you are anything like me, you will be watching with a sense of awe at the rapid advances machine intelligence is making as a predictive model across many industries. And it will come as no surprise that Facebook, Google ,Amazon , Tencent and Alibaba have invested heavily in ‘AI as a service’ platform capabilities. Companies that build commercial platforms with tons of data will be, and are, well positioned to provide insight based services that others will find hard to match. In insurance we can see a plethora of commercial use cases emerging including car damage assessment, managing fraud and abuse in health claims, drone-based agricultural algorithms, and chatbots galore.
So, AI is an undoubted technological juggernaut – and it has only just started to move. What may currently be used as second opinion tools, will be relied upon in the near future as primary diagnostic and advisory tools. This creates major implications for how humans re-skill themselves. Furthermore, the accumulation of technological benefits in the hands of the few is skewing wealth even faster globally and as a result is already leading to active political debate about the future of transformative technologies. Perhaps Elon Musk is right when he says AI is a threat to us, and we will need to work out how to neurally interface with computers in order to maintain a sustainable competitive advantage. In addition, we also see some moves to modify behaviour in China via social credit scoring – further details here
Setting aside these wider issues are some key questions of relevance to insurance. Instead of worrying about labour to capital substitution, or marvelling about the current and future use cases, let’s concentrate on issues emerging around Intellectual property and data ownership.
Data as a two-way value exchange?
Let’s start with a subject that is relevant for all of us – personal data, and let’s look at health insurance as here things get interesting quite quickly.
The first question to ask is who owns the data? In Europe regulators have made a clear policy decision that the consumer should own the data and be able to control the release and storage of their own data. But in health it is not so simple.
- What happens in a life-threatening situation where the patient is unable to give consent to medical professionals? Clearly as personal medical records become more essential we will need ‘intervention mandates’ to be structured so that medical professionals can quickly access an individual’s data in an emergency.
- What about the aggregation of health data at a meta level? This aggregation is also giving enormous insights and medical advances. If individuals are given the right to opt out then this risks the health of others. If individuals are given the right to be anonymous they are denying themselves potentially life-saving insights in the future. For example, if machine learning evolves, as it is doing already, too identify early indicators of diseases, such as Alzheimer’s, then the individual who contributed anonymously would not be able to benefits from early advisory services.
So personal data privacy will need to evolve beyond being 100% under the control of the individual. I would suggest there needs to be three additional components – emergency intervention protocols, social benefit sharing and mandatory health advisory. Of these the last one is hardest and the discussion should be based around merits of opt-in or opt-out protocols.
Now let’s move onto the IP around ‘insights’. On the surface we can all superficially agree with a statement that an insurer which gets insights from the data it manages is entitled to own that IP and profit from that IP. But is it really that simple?
What obligation should an insurer have to share insights on fraudulent claims with the rest of the industry? Is such sharing anti-competitive or good business?
With health insurance, what about the medical insights that insurers generate? Should they be obliged to share them with national policy makers? Can they charge a fee for these insights?
What happens when medical data and claims data is combined to create AI based insights around preventative healthcare? Is this IP that can be protected? The logic for drug patents has been based on the high cost of development of drugs but can this argument be used in this case? Should governments legislate and ensure that all AI medical insights are made available to all?
In the appendix I have outlined some of the IP ownership issues that the health insurance and healthcare industries will need to solve.
These questions of Intellectual property and data privacy will become more urgent to resolve as the value and impact of data continues to grow exponentially. Global bodies and national governments will need to co-ordinate as the private sector alone will not solve the problems and has economic incentives that suggest, if left alone, will be sub optimal for society. Data protection and the ownership of insights from the data are going to be hot topics of conversation and will increasingly be seen through the lens of human rights as well as through the lens of corporate IP protection.
Appendix: Health insurance in the not so distant future – a world of AI bots?
The Table below summarizes the world of bots that is likely to emerge soon and the IP considerations.
|Bot category||Example of use cases||Ownership of insights|
|Insurance company bots||Bots for fraud, waste and abuse||The company owns it|
|Industry bots||Data insight and AI insights generated from aggregating data across multiple insurance companies||Frameworks will be needed to manage these insight on behalf of the industry and also to share with national bodies for medical insights|
|Private medical bots||Specialist interaction to consumers e.g. chat based diagnosis tools, disease specific prevention tools||How should knowledge and insights that reduce medical risk and improve health outcomes be licensed?|
|National medical bots||Epidemic disease prediction, preventative medic||Owned by nations and global bodies for the benefit of society|