Open Response To a Shared-Life Operating System
9 January 2020
Dear Sebastian,
Earlier this week you posted a bold and energetic open letter on LinkedIn in which you described the concept of a Shared-Life Operating System. Your post was optimistic and I think ultimately correct about the direction that a number of our biggest digital companies -- Facebook, Amazon, Apple, and Google -- are not only likely to go but probably have to go in order to stay viable for the next decade. Let's call it your 20/20 vision of 2020 and beyond.
We have had conversations on this and related topics over the years. You know I respect your thinking and wanted to take this chance to respond to the details of your proposal as well as inject some additional thoughts. Part of this comes from my perspective as a consumer psychologist with a penchant for evolutionary psychology. I am always looking underneath the big movements of business and society to see why we evolved the desire or ability to do these thoroughly modern things in the first place. Because if a Shared-Life Operating System will work, it will only do so if it meets the fundamental, evolved needs of many of us individual consumers or users. Or, I should say more intentionally, us humans.
Because you have proposed your points in an open letter, I'm sending my reply as an open letter as well. I hope we can convince others to engage these points because I'd love to see what letters they would write back to us both. With that as my goal, let me lay out my thoughts under two specific headings:
- A Shared-Life Operating System is the most effective way forward.
- But is a Shared-Life Operating System the "right" way forward?
1) A Shared-Life Operating System is the most effective way forward.
The history of human experience was, at least until the dawn of the industrial revolution, one in which efficiency and effectiveness were applied only to a very narrow set of concerns, at least for most people not enmeshed in the Egyptian courts or serving in the bureaucratic machinery of the Ming Dynasty. Even in those highly dynamic systems, however, dramatic shifts in general wellbeing were infrequent and not always widely distributed. For example, if you were a Trobriand islander two hundred years ago, the only long-term enterprise you had to concern yourself with was the harvesting of tuber roots and the meticulous planting of the tubers in the next growing season, relying on ancient, collective wisdom to ensure that the roots would grow and that the spirits would bless the roots by sending enough rain but not so much that it would flood out the fields. This was no simple endeavor, don't get me wrong, but as countless anthropologists have shown, people who lived by hunting and gathering or by simple cultivation as the Trobrianders did had a smaller range of truly uncertain or variable concerns to address in a typical year than many of us moderns do in a typical week or even day. To be effective in their world required little in the way of data collection and few specialized tools.
Today, the potential to improve our outcomes is vastly expanded in contrast with any pre-industrial society. The combination of modern science and technological innovation has given us the ability to collect more data about cause-and-effect relationships and to communicate our analysis of those causes with increasing detail so that we can increase the good outcomes while reducing the bad ones. It's not an automatically efficient process nor is it linear in that good doesn't automatically flow from this process in a steady stream. But it is a process that can suggest more ways to improve our lives more surely and more consistently than ever before.
But it is a process that has reached a logical limit. You see, until recently, there were still many independent domains of our life that we could improve individually without worrying so much about how they overlapped. This made these disparate domains simple to approach and address. We found that iodized salt improved mental development so we could encourage or even mandate salt iodization. The number of independent suggestions that have profound effects on human flourishing has crescendoed to a magnificent degree. Flouride in the water, awareness that tobacco causes cancer, the polio vaccine, blood donation and portability, dialysis -- the list of independent ways we can improve our individual and collective lives through science, engineering, and technological innovation would occupy as many pages as you could care to read. We can and should continue to invest in these types of innovations. But we are finding the limits of the single-cause, independent interventions we can propose to improve human wellbeing. Instead we are entering an era of mostly polycausal phenomena.
Want to know what causes heart disease? There is no single gene, gut bacterium, food additive, environmental toxin, behavioral intervention, or even medication that can influence heart disease reliably. Not for any one single individual and certainly not for humanity as a whole. Instead, we have to prepare ourselves to approach heart disease as it really is, as a polycausal phenomenon that is caused by -- or prevented by -- a multitude of factors that each interact with each other to varying degrees, all in ways that we can't predict with our current understanding and the data available to us. That we keep approaching things like heart disease with single-factor or monocausal solutions -- prescribing everyone at risk a daily statin even though the statistics of statin effectiveness suggest most people who take them are not likely to benefit at all and some are even harmed -- is a testament to how well we learned the lessons of history but how maladaptive this learning now is.
Those historical lessons now apply to a diminishing number of human endeavors. This is not a surprise to anyone who has done research in any field in the past 50 years -- my own work studying the effects of television in the 1990s underscored for me something that prior media researchers had already learned but were hesitant to discuss publicly, that TV was a huge thing in people's lives to be sure, but which had as many possible effects on people's lives and could be used for as many possible gratifications as there are people. The same is true today for the effect of social media on depression, the role of Internet shopping on the climate, and the ultimate cause of the troubling rise of suicidality among young people in the developed world.
As humans, we continue to aim for single-factor solutions because that's how our minds work. And they worked very well for a world where we were planting tubers or hunting bison in our clans of 20-30 close kin. But they don't work very well when we have to consider the many variables at play in polycausal phenomena. Some variables we can't even see, others we can see but can't measure, and others we can measure but can't rightly understand, often because we can't see how they interact with the other variables we don't see or can't measure. Resolving this with our limited minds would seem hopeless.
Except for one thing: We stand on the edge of a future where machine intelligence will be able to detect, deduce, measure, and analyze the variables we tell it to as well as the variables we don't know exist. Machine intelligence will soon be able to tell us what combination of factors we should add, maintain, or subtract in our lives to prevent heart disease reliably; to reduce depression incidence or at least cope with its devastating presence; and to strengthen the relationships that matter to us.
I won't take the time to explain how this is technically possible. I will instead focus on the requirements for making this happen. The biggest requirement is data. To help me pinpoint my actual risk of heart disease years before it would manifest in a traditional physical exam, the system would have to know my genetic makeup -- and not just mine but everybody else's so it could discriminate which alleles are likely to increase or decrease my risk based on the outcomes of others. The system would have to have access to my current biometrics including things like hormone functioning, immune response performance, nutrition profile, and so on. Here again the system would need to know the same data from millions of others about whom it would also know their genes and their health outcomes so it could model which causes are likely to have effects and conduct natural mini-experiments on people who are proxies for me to see which behavioral, nutritional, medical and emotional interventions are likely to have the most dramatic effect. The system would also have to understand my emotional and motivational states so that it could select among interventions that I am most likely to successfully accept and implement, crafting the right message to me that resonates with the way I see myself and my world.
You can see how complicated this gets and yet I'm only talking about heart disease. What about strengthening my personal relationships? Raising an emotionally robust child? Performing well at work? Choosing a hobby? Determining how much I can volunteer in my community? The number of decision outcomes that we could improve in our lives is vast. Remember: Self-help books sit alongside cookbooks as the perennial money makers in the publishing world. That's because people always know their lives could be better but aren't exactly sure how to make them better. And the common approach -- imagining that each of these things could be improved with a single-factor intervention whether it be the Marie Kondo method or joining Peloton -- doesn't consistently deliver what we hoped.
In your letter on Share of Life, you propose to brands that they recognize this yearning that consumers have and that they engage with or entangle with their consumers to consistently and reliably deliver meaningful outcomes to them. I applaud that and believe it's a good strategy. But as you make clear, your approach to Share of Life opens the door to a much deeper entanglement that brands typically imagine they can or should approach. To get even close to that, brands would have to know much more about their customers than they do today. That data can't sit in silos where one brand knows about my steps, another knows what food I buy or eat, and yet another knows my medical record or how I interact with my children. That data has to be brought together into the polycausal inference machine that the Shared-Life Operating System will have at its core.
What entities -- brands, companies, cooperatives, governments -- will ask for and receive permission to have this data and use it on your behalf? In your model the most likely contenders are the Fab Four. Apple is already using Apple Watch Series 5 to save lives via real-time EKG monitoring. Google can already make all the decisions you need to get you up, have you dressed for the right weather conditions, departing your home to deal with current levels of traffic, and routing you to work in the least-stressful way possible. Is it really such a stretch to imagine that these companies might very soon ask for permission to help you keep your heart healthy or your marriage in tact?
2) But is a Shared-Life Operating System the "right" way forward?
In the 1420s, more than a half century ahead of Christopher Columbus's voyages across the Atlantic, the Ming Dynasty was sending fleets of massive trading ships up and down its side of the Pacific and as far west as Arabia and the eastern coast of Africa. Though accounts are only somewhat reliable, it is likely that this fleet was the largest in the world at the time and the source of unprecedented wealth and international trade. Yet, in 1425, a newly ascendant emperor to the throne overturned his father's legacy and ordered the end of the sailing fleets. Different accounts suggest that the ships were burned and some indicate that the shipmasters who built them were executed, all to prevent further journeys like those undertaken previously.
We don't have reliable accounts of the thought processes involved, but one thing we can safely surmise is that the powers that be looked at the bold, new world opened to them by the sea and simply decided that they were having none of it.
The moral of the story is obvious for today. For a variety of reasons, some of the people who read what I've described above about the Shared-Life Operating System and its potential power to dramatically improve human wellbeing will shrink and prefer to burn the boats and jail the shipmasters or worse. You've anticipated this in your open letter where you talk about the push to break up big tech -- a hot topic in the US after the 2016 election and one that will only get hotter.
I can't answer for everyone else. I understand the psychological and emotional factors that would hold someone back from supporting the next steps I've outlined for human flourishing. Some distrust systems they don't understand, which is rational. Even if we do understand the system in principle, I am all too aware that human organizations -- businesses or governments -- are fallible and prone to inefficiency and outright manipulation. Each of us has a set of interests we would prefer to see running these companies or governments and we won't agree between us on which interests those should be. So it seems impossible to come to a conclusion about how we make the move from a monocausal, single-factor approach to human wellbeing to a more sophisticated, polycausal approach. It might be easier to ban the boat and stick the land we know.
But to answer my own question above -- is this the "right" thing to do -- I want to first propose that even if we aren't ready to support it, we should at least not conclude that it is the "wrong" thing to do. It may sounds like I'm parsing this too carefully, but in my work I encounter many people who have already come to that conclusion. If my experience is of any benefit here, I owe it to us all to share that I have decades of consumer psychology data that tells me that people who immediately reject an innovation are rarely the Cassandras they imagine themselves to be -- prescient seers who should get to dictate the possible solutions the rest of us have available to us. If they were correct, the many things those people were afraid of would have collapsed by now, serious things like in vitro fertilization as well as the more quotidian things like online shopping.
Yet I can't pretend that the answer is to simply say, "If you don't want to enter a Share of Life relationship with a company, that's fine, you do what you want and I'll do what I want." This is partly because some of those people want to legally prevent companies from offering a Shared-Life Operating System to help me. But it's also because the system is powered by a network effect. The more people who contribute, the more accurate the recommendations will be, the more we can measure the outcomes against predictions and adjust them. There's also a long-term moral concern on the horizon: People who choose to be left out today will not be figured in to the algorithms. It's very likely that the same personality profiles that don't want to be included will have unique needs or benefit from unique interventions. At some point we want to find a way to make the benefits valuable to those people so that they can be included in the system.
Yet, as a colleague recently put it to me, by encouraging the development of this system, I'm depriving her of the ability to opt out of the system, effectively restricting her agency. Because once the ecosystem to support this more effective approach is built up, the incentive to maintain parallel ecosystems that support objectors will be reduced. Just as we saw the hyper efficiency of Walmart diminish the incentive or ability for small-town retailers to stay in business. This is true, I have no answer for that, other than to suggest that if we step forward into the future that you and I are collectively envisioning, the more openly we do it, applying the principles of transparency and customer control that are emerging from today's personal privacy debates, the more likely it is that the Shared-Life Operating System you propose will be not only effective, but ethical.
I don't claim this brief discussion answers the objections to so much data collection and analysis. In fact, answering the objections is not possible or even desirable. No human endeavor from the cultivation of penicillin to the moon landing has ever proceeded from origin to glory without a mix of detractors and supporters, oblivious cheerleaders and blind recalcitrants. To my thinking, this is one explanation for our collective success -- our vast range of human impulses combine in various ways, acting like a massive meta-brain to help us coordinate tasks far too complex for one person or one group to approach or accomplish. By taking advantage of this mind we have already accomplished many complicated tasks. Once we add machine intelligence to this process, powering the combination with a Shared-Life Operating System inference engine like the one we're describing, our ability to increase human flourishing will expand and accelerate. The waters are choppy but it's time to get in that boat and start sailing!
Best,
James McQuivey, Ph.D.