Equideum Health (formerly ConsenSys Health) has partnered with Nokia Bell Labs to research cognitive augmentation using blockchain combined with AI and edge devices such as earpieces, smartwatches, and home devices.
The idea of Elon Musk’s Neuralink brain chips springs to mind with the mention of cognitive augmentation. What Equideum Health and Nokia Bell Labs plan to explore is quite different. Imagine that someone is cognitively impaired, whether because of dementia, an accident, or a congenital disability.
Nokia has developed a wearable ear device – an ‘earable’ – with a microphone and speaker that can use on-device machine learning to help people understand and respond to the world around them.
Clearly, this kind of device receives a broad range of very private data, raising various ethical issues around the data.
And that’s a key driver behind yesterday’s announcement. These and other IoT medical devices are developed and rolled out, but the data management infrastructure is often lacking. That means that either it’s hard to access the data for machine learning, or there’s a risk of exposing patient data.
“The announcement with Nokia Bell Labs today is about a joint applied research agenda on how we will approach the ethics, the identity, the cybersecurity, the compliance challenges with respect to edge data on all devices, including earables, but every other remote telemetry device currently being used,” said Heather Flannery, CEO of Equideum Health.
And this applied research phase will not use real personal data.
Data doesn’t need to be shared to be useful
Modern clinical research uses machine learning which needs data and lots of it. Even if it’s a simple wearable device that monitors your heart rate, that data could be useful without sharing it. In federated learning, a program could analyze the data on the device and deduce a result. And the result could be used for machine learning. That could be as simple as using your age range and daily heart rate range. But it could also be a little more private such as you have diabetes, but not your name.
The alternative is to share the underlying data but in a privacy-preserving manner. Both of these approaches will be explored during the research, in all cases, only with the patient’s consent.
Monetizing the data
Last year, Nokia launched a blockchain-powered Nokia Data Marketplace to enable data to be sold and used for AI and machine learning. That version was based on enterprise blockchain Hyperledger Fabric and did not relate to health.
In a separate agreement from the earable research, Nokia and Equideum partnered for the Equideum Exchange last month. It’s similar to the Nokia Data Marketplace, but this time using a public-permissioned Ethereum blockchain architecture for health data.
US requires legislation that clinical providers, insurers and pharmacies share data with the patients if they ask. Hence Equideum Health is developing an infrastructure to enable patients to store their own electronic health records (EHR), which they can get from their providers. They then will have the option to monetize the data in their EHR, for federated learning or share the underlying data with consent.
The platform supports self-sovereign identity, privacy preservation and also uses decentralized finance (DeFi) to monetize health information.
Stepping back, the availability of mass data could result in massive leaps in healthcare through machine learning. However, if granular machine learning results are shared with insurers, that could result in insurance coverage being less affordable for those most in need, as explained here.
Provided people know what they’re getting into and there aren’t surprises, individuals should be free to choose whether or not to monetize their data.
However, the choice is a bigger issue where a government (or insurers) can potentially access the data or the patient has a cognitive disability.
Flannery was keen to emphasize that these issues exist today. And in fact, its solutions aim to address them. “We at Equideum are frankly civil libertarians and radical privacy advocates. We seek to implement empowerment systems for individuals that balance asymmetry and power between the governments and enterprises that seek to serve them,” said Flannery. She emphasized the potential for federated learning without sharing underlying data.
Regarding disabilities, the concern is whether a guardian or legal representative might opt-in to share data that the patient would not agree to if they were capable of making the decision themself. That’s especially where financial incentives are involved.
“We believe we bring improvements in transparency, auditability, privacy, cybersecurity. Improvements over baseline versus further infringement or compromise on the agency and potential of the individual,” said Flannery.
There have been many press stories of people in care who have been mistreated or neglected. Beyond wearables, most care homes have sensors and monitors, which are also IoT devices. Dr. Pierre Vigilance, who heads up the Equideum Foundation, noted that the management of that data could also help to ensure people are better cared for.
One of the challenges with sophisticated technologies is that consumers don’t understand them, resulting in surprises. Two years ago, Equideum Health was spun out of ConsenSys, which operates the Metamask cryptocurrency wallet. There’s a perception that self-custody wallets are private compared to transacting via cryptocurrency exchanges. However, in February, many were surprised to discover their transactions don’t go directly from Metamask to the Ethereum blockchain but pass through centralized ConsenSys Infura infrastructure, which monitors the IP address and can use it for sanctions.
Another example relates to federated learning. While it may not share underlying data, there are risks of data leakage.
Despite the risks, we shouldn’t lose sight of the potential benefits of greater data sharing, ultimately saving lives and achieving better health outcomes. Envisage a cognitively impaired person with an earable device that helps them feel less confused.