Your eyes will be discontinued: what are the long-term responsibilities for implants?

by Anders Sandberg

What do you do when your bionic eyes suddenly become unsupported and you go blind again? Eliza Strickland and Mark Harris have an excellent article in IEEE Spectrum about the problems caused when the bionics company Second Sight got into economic trouble. Patients with their Argus II eyes found that upgrades could not be made and broken devices not replaced. What kind of  responsibility does a company have for the continued function of devices that become part of people?

In many ways this mess is unsurprising: we see similar “bricking” (where unsupported technology doesn’t just stop getting updated but stops working altogether) of many other “smart” technologies, whether home automation, games, railroads, or robots. As Jeffrey van Kamp said in his eulogy to his dying Jibo robot: “Nothing is permanent if it requires a server to operate”. Many of our technologies are densely interconnected with external systems that require somebody to pay for their server space and maintenance. When a company goes bankrupt, changes business, merges, or no longer wants to support older systems, they will fail.

Obsolete implants are also a medical issue. Implants do fail over time. Some people go with several old pacemaker leads in their heart, and repairing old dental or hip implants when their maker is unknown can be challenging. Some implants may also make further medical interventions harder, like the bionic eyes making neurologists wary of performing a MRI scan.

A part of oneself

What makes the Argus II eye case more ethically intense and interesting is the close link to the user.

The bionic eyes are senses that performs a function in their lifeworld. A functional neuroprosthetic becomes a part of the sense-cognition-action feedback loop. This is not magic plug-and-play by any means: training how to use the implants takes time and effort. During this process the brain learns to interpret the signals, and the patient learns behaviors make use of the prosthetic. At least artificial limbs appear to become encoded as real limbs in the brain; research on artificial retina representations seem to be lacking, but given the flexibility of body representations it is plausible that they too can become part of the sensory-body schema, not just as external tools.

In a real sense the system is part of their self just as an ordinary body part is. We might hence make the same move as Chalmers and Clark do in the extended mind thesis: if something is doing the same work as we normally do with our body and forms a tightly coupled system with us, we should regard that as part of our body. Mary Jean Walker argues on several grounds that “bodies’ normative status relies on the relation of a body to a person and shows that persons could have similar relations to prostheses”. Sven Ove Hansson reaches the same conclusion, noting that this close link makes explantation, the removal of implants, ethically potentially problematic.

Ending clinical trials of implants can not just leave participants with non-functional implants in need of removal, but also a loss of function. Many brain-implant trials do not cover costs of explantation or new batteries to maintain function. The Dobelle project, an early trial with cortical artificial vision 1975-2005, is a good example in this context: participants that gained sight felt a profound sense of loss when they were deprived of these abilities due to device failure or the termination of the study. In this case the situation was made worse by subjects having to pay up to $200,000 for participation – they had themselves contributed significantly. As noted by many ethicists this is not OK.

Joseph Fins has argued that the neuromodulation community should adopt an ethical principle of non-abandonment. Investigators and sponsors “incur a clinical responsibility to provide on-going care and a fiscal responsibility for any associated costs. It is a breach of professional ethics to do otherwise”.

What to do?

Unfortunately, saying the company morally has to support implants leads nowhere: if it goes bust it cannot do it. Ought implies can, cannot implies no ought.

Adding after-care provisions might work for companies changing business or merging (merely making the implant market full of liabilities, slowing development and deployment), but this likely works better for explantation than maintaining a function. While current implants have a limited external infrastructure (charging, interfaces, protocols for programming) future implants are likely to have even more external complexity – with the same issues as we have seen with other “smart” devices.

Saying that it is all late stage capitalism also leads nowhere: while a state-owned medical system is less likely to go bust, programs and funding do end. Plus, usually health systems do not make the implants anyway, and are presumably unready to take on more complex responsibilities if a company cannot do it.

Saying that we should not focus on medical technologies in need of sophisticated maintenance also leads nowhere: usually the reason they are developed is that there is no easier treatment. Refusing to develop such devices on the grounds that they might be hard to maintain leaves the patients untreated.

Probably one way forward is to argue the right to repair applies extra strongly here: technical and software information must be made available. This might be something one could enforce legally, either by initial contract (“In case of provider bankruptcy…”) or by introducing laws mandating that exoself services must be continuable. This is still a limited improvement: maintaining a complex device may still be too expensive for individuals or patient associations.

Probably we need to strongly push for open APIs and other open-[somethings] when it comes to medical technology that can last longer than companies or organizations. An interesting example is the development of an open hardware and open software visual neuroprosthetic (github repository), although at present it is only intended for animal experiments. Making this kind of open systems work with corporate IP and elaborate medical device regulations will be a nontrivial challenge, but likely morally worth it.

Active neuroprosthetics become a part of the person through a process of adaptation and training, but as demonstrated by the stories of the Argus II (and Dobelle) users patients also take an active role in literally building themselves. The whole discontinuation issue shows that ethically there is a need to have informed consent not just about the medical functionality but also the wider economical-technical functionality and actual autonomy about the function. If one cannot control one’s body parts – natural or artificial – one does not control oneself.