Prototype_ Care AI
#Health | #IdentityManagement | #Authentication | #Anonymisation | #ArtificialIntelligence | #SmartContracts | #InvisiblePopulations
Narrative [7-8 min read]
Care AI belongs to the health sector. The group behind it includes Gui Seiz (FabLab Barcelona, IAAC / Institute for Advanced Architecture of Catalonia) and Jordi Planas (Vimod Studio) as lead designers, Maciej Hirsz (Parity), Ivo Lõhmus (Guardtime) and Annalisa Pelizza (University of Twente) as expert stakeholders, and Lucas Peña (Ideas for Change) in support to the prototype production.
The main concern behind Care AI’s development was the existence of invisible populations who are unregistered in traditional healthcare systems. Sometimes they are travellers or professionals temporally working in other countries, but most remarkably they are deprived social groups such as migrant refugees or people in situation of homelessness. These populations experience constraints in accessing healthcare due to legal, social or economic limitations, and usually find themselves in the absence of answers for their situation.
Care AI is a platform conceived to allow such invisible populations not only to get access to basic healthcare, but also to information about other means by which they can obtain support to cure their illnesses and ensure a sustained good quality of life. What is more, if for some of these groups the crucial point is the need to access healthcare without compromising their identity, Care AI is designed to facilitate it.
The Care AI prototype was developed around questions of identity management and data authentication and certification, and its conceptualization orbited around data anonymization, or at least data pseudonymization. According to the assumptions of the group behind Care AI, sensitive health data would not be directly linked to personal identities, but to elements such as anonymous cards and registries. So compliance with data privacy regulations would be more easily achievable, even if further legal analysis would be required considering new data security policies and regulatory frameworks such as GDPR.
While personal identification does not follow established legal and jurisdictional definitions in the Care AI scenario, an alternative understanding of “identity” based on health data is proposed. As a result, the proposed solution allows the inclusion of people usually left out of healthcare for diverse reasons, but doesn’t create further exclusion.
The Care AI system operates as a network of micro-entrepreneurial owned, automated Care AI Points. Each Care AI Point provides non-hospitalized test and diagnosis to people without access to traditional healthcare. The Care AI Point interacts with a Care AI Smart Contract running on any smart contract enabled blockchain, such as those possible through the Ethereum network.
Within the scenario proposed by Care AI, we find Erin, an undocumented migrant in need of medical assistance for an illness she contracted recently. Due to her irregular migratory situation, Erin is afraid to visit a doctor within the national healthcare system, and cannot afford any of the private care services offered at her location. Through the implementation of Care AI as a service targeted to individuals with her profile, Labs-on-Chips (LOCs) are distributed for free in homeless shelters, pharmacies and other places, and grant access to any Care AI Point.
Erin acquires a LOC in one of these places and finds out that it includes a fingerstick for venous blood collection and analysis. She reads the instructions provided, pricks her finger and deposits a small sample of blood onto the chip. Erin proceeds to enter the chip into one of the Care AI Points, which establishes the connection between her blood sample and her identity with an anonymous card.
A recurring user would be able to scan their card at any Care AI Point. New users would be invited to generate new private keys and get new cards printed during first interactions with the system. However, the personal identity card used in Care AI could have several categories of personal data. It could even be one that users already possess for other services, provided such services offer means for interaction and authentication that are adequate for anonymization.
In the prototype, this card is simulated through a generic QR code card. But in a real life scenario it would also include a private key as simple means of authentication to the system, decrypting the medical history of the card holder and uploading new encrypted records onto the Care AI Smart Contract.
Erin’s blood sample is anonymous and analysed at one of the Care AI Points by a HealthBot assistant. It asks for consent for sharing this data for research and upon approval, it prints a receipt with a potential diagnosis and suggestions for further action.
This action may depend on the degree of confidence of the analysis, but it may go from recommendations for self-care, to prescriptions at participating pharmacies, and escalation to medical attention at NGO doctors. If the process goes accordingly to previewed, Erin would then be asked if she would like to add the results to her anonymous logbook for later reference. If so, another card with a bespoke QR code would be issued, which then Erin could use later on to create a medical issue with each data exchange with a Care AI point.
The underlying idea behind Care AI is that in exchange for the analysis, suggestions for further actions or additional information, the person also donates its anonymized health data. And in all this, Care AI strives to promote the creation of anonymized medical history for future diagnosis and potential integration into traditional healthcare systems.
However, the amount and type of information the user decides to provide to the Care AI system can vary greatly. In this scenario, variations would also imply different gains for the user in a sort of credit system, whether in terms of the information received, whether in terms of possibilities of access to healthcare.
Public authorities could be the first actor managing this information, and letting the system to self-finance itself through the creation of distributed apps (Ðapps). These Dapps would allow not only public health and research bodies to access the anonymised data through smart contracts inscribed in the blockckain, but also market-oriented third parties to make use of the same information if granted access.
Payments would serve to subsidise the medical treatment while also paying the creators, owners and maintainers of Care AI Points. Diagnostics work is delegated to an AI, but this is not expected to cause a loss of medical jobs, considering the user would not access healthcare without Care AI and NGOs being involved in case of escalation.
Access to medical insights could help to better plan public funding and policies, such as in the forecast and management of seasonal outbreaks. But the potential of the data collected and exchanged through Care AI could even go beyond health. It could allow authorities to become aware of other ongoing issues within these invisible communities, thus building information on socio-economic dimensions or demographic shifts, and plan accordingly in an integrated fashion.
Other actors could come also into play in the same scenario, such as startups and other SMEs. Having heard of the Care AI project, they could download Care AI Point blueprints, fabricate their own or even iterate new models. Bound to the main smart contract protocols they would be only dependant on public authorizations to deploy the Care AI Points into the public space. These companies would receive remuneration for accessed data, and as such would be incentivised to keep their machines working.
In addition, to foster the growth of the network, a research spin-off that provides LOCs or works with Care AI Points could run Initial Coin Offerings (ICO) to fund the research and development of the Care AI Points’ open hardware and software specifications, the smart contract code and also the data specifications.
Questions may emerge, however, with such new business and innovation models. These third party actors can exploit low entrance barriers in producing LOCs and replicating Care AI Points for instance, and flood the market in ways that are counterproductive considering the main social goals of Care AI. Or yet, big and established biotech companies can take advantage of their forefront market position and become the major supply of LOCs to all micro-companies deploying Care AI Points, thus promoting monopolist practices.
In this case, a central body could issue a payment to the Care AI Smart Contract, which would then allow it to decrypt a number of records collected by different Care AI Points, in a first-in first-out manner. This could help to redistribute the payment not only to managers or maintainers of different Care AI Points, but also to different providers of supplementary technologies such as LOCs. Moreover, it could also help not only to cover costs and provide economic incentives for a bigger number of players, but also the creation and inputing of soft regulatory measures into the market.
Beyond its main social function, the Care AI prototype aims to showcase how to stimulate entrepreneurship through open hardware and distributed manufacturing. It would create a blockchain based marketplace in which the local Care AI Point providers, parties interested in purchasing the anonymous medical history records, and healthcare authorities could cooperate in a trust-free fashion.
But looming questions remain in the end. Does all this respect or even consider ethical implications of a personal-data-for-healthcare exchange? And will any of the players involved in the production and exploitation of Care AI Points ever think about the kind of problems they might be causing for the creation of more inclusive healthcare systems?
Additional Materials for Download>
Care AI and its respective text, design, audiovisual, and code elements are made available by the EU Policy Lab of the Joint Research Centre under a EU Public Licence (EUPL), and where not applicable under a Creative Commons Attribution-Share Alike 4.0 International (CC BY-SA 4.0), unless otherwise stated.
You can use, redistribute and/or modify the whole or parts of the available material under these terms “as-is”, as long as you note the licenses and attribute credit (European Commission, Joint Research Centre, lead designers, other contributors). This is without warranty and declared fitness for any particular purpose, and disclaiming liability for any consequences resulting from using, redistributing and/or modifying any of the provided elements in any given circumstances.Flyer Source Files (.ino + .py)