10 Coming Together

Let’s revisit Vögisheim, the quaint village in the southwest corner of Germany where our journey (and my life) began. Many locals and visitors call it the Tuscany of Germany. My Italian friends might consider this a bit of an insult, but the region does bear some resemblance to the birthplace of the Italian Renaissance. Rolling hills and vineyards as far as the eye can see. Rows and rows of grapevines dividing the hills into symmetrical parcels. The regions’ wines aren’t as well-known as Chianti or Sangiovese, but if you are a wine connoisseur, you might recognize our Gutedel or Spätburgunder.

When I was growing up, my parents still owned vineyards, even though my grandma’s family gave up farming in the late 1950s. Every fall, around mid-September, we harvested.

Harvest season was fun (at least for the kids). Armed with big grape hoppers (large backpacks used to carry the grapes) and a sharp pair of garden scissors, we’d walk up and down the rows of vines, snipping and tossing grapes over our shoulders into the hopper. Once filled, we’d empty them into a trailer hitched to a tractor.

As kids, we probably did that for five minutes to feel important before we ran off to play in the fields, watching the adults labor away. At lunch everybody would sit together and eat homemade bread, cheese, and cold cuts. I should say that this entire experience became a lot less romantic once we were teenagers and expected to pull our weight.

Whenever I’m back home, I go for a walk in the vineyards—no matter the weather. One time, I brought a friend visiting from Japan. We walked through my family’s vineyards, and I told him about my childhood adventures. He wanted to try our wine. I can imagine you might too.

I had to giggle a little. Our vineyards are far too small to produce our own wine. We neither had the equipment nor the expertise to do so. But, looking back, his question was a good one. He wanted to know, “What happens to the grapes?”

Selling the grapes to one of the big wineries in the area could have been a possibility. But we never harvested enough for this to be a real option. Even if we did, any deal would have greatly favored the winery. So instead of simply letting the grapes go to waste, most families in Vögisheim and the surrounding villages were part of a Winzergenossenschaft. Yes, a true beauty of a German word. The English translation is winemakers’ co-op.

After harvesting the grapes, we dropped them off at the co-op. The co-op would either turn them into wine or sell them to the wineries.

Working with these co-ops had several advantages. First, combining the grapes increased the value of each individual harvest. Second, the co-op brought expertise, both because many of the members were winemakers and because we could pool our resources. The proceeds from the wine and grape sales allowed the co-op to buy advanced equipment, hire expert oenologists to improve the quality of the wine, and bring on marketing professionals to sell it. More than any one of us could have pulled off alone. Pooling our grapes made all of us better off.

The same is true of your personal data. Your individual data isn’t worth very much. It only becomes valuable when combined with the data of others. Think of medical research. Your medical record alone won’t tell us anything about the risk factors associated with a certain disease. We can only start exploring these factors once we have a sufficiently large pool of carriers (and noncarriers).

The same is true for your Facebook and Google data. The two companies only care about your data because they can connect and compare it to the data of millions of people. That’s what allows them to extract the insights third parties are willing to pay for.

But it’s not just the value of your data that increases when you pool it with others. Just as my family didn’t have the expertise to turn our grapes into wine, most of us don’t have the expertise to make good decisions regarding our data (see chapter 8). Left to our own devices, we simply don’t stand a chance. We have neither the expertise nor the time. Could my parents have figured out how to make wine? Probably, even though it might not have been very good. But were they eager to dedicate their whole lives to this? Hell no.

Just as the people in my village came together to reap the fruits of their labor, we need to come together in small communities of like-minded people to collectively manage our data and benefit from it. Like wine co-ops, data co-ops are member-owned organizations that pool and manage their members’ personal data to benefit the collective. However, unlike wine co-ops, data co-ops don’t require people to be in the same place—although they could. Instead, the members can be connected by a common goal and a shared strategy for leveraging their data to accomplish that goal.

Digital Data Villages

Let me give you an example of how a data co-op could work.

As I started writing this book, I got pregnant. A beautiful but also terrifying experience. You get advice from all directions. Do this. Do that. Most of the advice will at some point contradict prior advice you’ve gotten. Eating sushi might put the baby at risk. No, that’s not true. What you must look out for is caffeine. You have access to doctor check-ins every two to four weeks. But what you really want is a minute-by-minute update on how things are going, and the assurance that everything is fine. All this uncertainty drove me nuts.

Now, imagine expectant mothers from around the world sharing their genetic and biometric data, alongside information about their own health and the health of the baby. You could stop the guessing game, and instead base your decisions on actual data. To start with, you could build advanced predictive models to identify general risk factors. Some of them might be known already, but some of them might be new.

Not just that, the members of the co-op could receive personalized, dynamic predictions of their own risk factors and current pregnancy status. Or customized advice on how to cope with morning sickness (a very misleading branding for all-day misery) or the constant fatigue.

By tapping into different data sources, the model could form a holistic impression of the mother’s circumstances. Who is she (e.g., age, ethnicity, historical health records, levels of physical activity)? What’s her social context like (e.g., Is she a single mom? Does she have a lot of support from other family members)? And what’s the potential impact of her environment (e.g., Does she live in an urban area with high levels of air pollution)? Combining all these factors, is there anything our expectant mother should be worried about? And if so, what should she do? I would have signed up for this data co-op in a heartbeat.

You could think of many, many more examples. I’ve only listed a few here:

What is common to all these examples is that the individuals involved voluntarily share a selection of their personal data with the co-op to help the entire co-op benefit from the group’s insights. Having access to my own genetic data is useless if I am trying to figure out how to improve my pregnancy experience or the health of my future child. But it could be extremely valuable when pooled with the genetic data of other expectant mothers.

Data co-ops turn the existing data model upside down. Instead of a few companies controlling and profiting from your data, you decide who to share your data with and you benefit from doing so. This works because data co-ops (and data trusts) are owned by their members and bear fiduciary responsibilities. They are legally obligated to act in the best interests of their members. And because co-ops are effectively governed by their members, anyone who joins the crew gains partial control over how the co-op is run. The system runs on collective rights and accountability, as opposed to exploitation and obfuscation.

This shift in the ownership and incentive model makes co-ops ideal champions for the privacy-preserving technologies I introduced in the previous chapter. Federated learning wasn’t developed specifically for data co-ops, but these organizations could be among its early adopters because they have a strong incentive to use such technology. This is unlike Facebook, which profits from accessing as much user data as possible. That’s its business model. Data co-ops are the exact opposite. They act on behalf of their members and are measured by how successful they are in amplifying the benefits and mitigating the risks.

The specific goals of any given data co-op can vary. For example, some co-ops might focus on helping individuals monetize their data. Much like our wine co-op enhanced the returns on our grapes, data co-ops boost your bargaining power. With 20 million allies, the big players will suddenly have to take you seriously.

But monetizing data is just one of many potential goals of data co-ops and perhaps the least compelling. Data offers insights that can improve our lives and those of others. I would not sell my medical records for additional cash in my bank account (fully realizing that this is a privileged position to be in). But I would happily give it to a trustworthy organization focused on improving the health and well-being of expectant mothers and their babies. For free.

Let’s look at some examples of what data co-ops might look like in practice. Although the concept is still relatively new, there are already several success stories.

There’s the Driver’s Seat Cooperative, a ride-hailing app that allows drivers to share their route data with one another and benefit from the collective insights this data can generate. What’s the fastest route for this delivery? What’s the best spot to pick up customers at 2 a.m.?

There’s also the Swash co-op, which pays its members for browsing the internet by aggregating and selling web activities in a privacy-preserving way (with full control of each member over what data is collected).

And then there’s my personal favorite: the Swiss data co-op MIDATA.

Changing the Swiss Health-Care Landscape

MIDATA was established as a nonprofit in 2015 by a group of scientists at ETH Zürich and the University of Applied Sciences in Bern (aka, my next-door village neighbors). The co-op acts as a trustee for its members, who can contribute to medical research and clinical studies by granting access to their personal health data on a case-by-case basis. Think of it like a bank account for health.

Anyone can open an account and deposit copies of their medical records or any type of health data that is valuable in this context (e.g., smartphone sensing data). MIDATA makes sure your data is securely stored in its collective vault and gives you full control over its use; you decide who to grant access, to which particular type of data, and for what purposes. And you can withdraw your personal data at any point in time.

But unlike your typical bank account, MIDATA isn’t interested in generating profits (no ludicrous late fees). Its sole purpose is to maximize value for you and its other members. Any net profits that are generated from the use of your data get reinvested into making the services on the platform better (i.e., advances in data protection and technological developments). Similarly, as a member of MIDATA, you aren’t just a bank customer. You own the bank. Literally. Control at MIDATA not only means control over your personal data. It also means having a direct say in the co-op’s governance through a general assembly (so Swiss!).

The value MIDATA generates for its members takes different shapes. You can share access to your data with third parties to improve your own health. For example, there are personalized applications to help members overcome addiction or fight obesity. But you can also share your data to support scientific discovery, for example, to help researchers better understand allergies, food sensitivities, or rare disease. In many cases, the same application does both.

Take MiTrendS, an application that drives the scientific exploration and personalized treatment of multiple sclerosis (MS). MS is a chronic autoimmune disease that affects the central nervous system and cannot be cured (yet). By eating away the protective layer of nerve fiber in the brain, spinal cord, and optic nerves, MS significantly impacts patients’ quality of life. They might have trouble seeing, experience fatigue, have a hard time concentrating and remembering things, and struggle with balance and tremor. A vicious combination of symptoms that often makes it difficult—if not impossible—for MS patients to thrive and fully engage in social life.

Although MS affects over 2.5 million people around the world, the disease remains hard to diagnose and even harder to treat. All we know is that MS is driven by a complex combination of infectious, genetic, and environmental factors. Because every patient has their own unique set of symptoms and factors that contribute to the outbreak, understanding the disease and developing targeted treatments requires large amounts of data. Not just data from a lot of patients, but also data from a lot of different sources: genetic data, medical history, exposure to environmental risk factors, medication, progression of symptoms over time, and more.

MiTrendS empowers patients and doctors to do just that. The app allows users to track their symptoms over time from the comfort of their home. For example, the app might ask you to follow a line on your tablet with your finger as quickly and exactly as possible to test your fine motor skills. Or match numbers to shapes to assess your attention and concentration. By combining these symptom assessments with existing patient records (i.e., medical records, medication information, brain scans, blood analyses, and more), MiTrendS can develop personalized treatments and care plans for each patient, a revolutionary approach that could change how MS is diagnosed and treated.

Of course, the big pharma companies might occasionally invest in large data collection efforts or buy patient data from hospitals to study the disease. MS medication is expensive, and there’s money to be made. For the pharma companies, that is, of course. Not for the patients whose data is used. In the best case, these patients will end up paying for the medication. In the worst case, they will never get to reap any benefits themselves.

The MiTrendS application turns this model upside down. By sharing their data, patients help develop better treatments for the future. But they also directly benefit from more customized and targeted treatment in the here and now.

Once a patient’s data is securely stored and combined on the MIDATA servers, a machine learning algorithm developed by researchers at ETH Zürich creates an optimal, personalized treatment plan for them. The algorithm not only considers the patient’s unique circumstances, but also leverages the insights obtained from all the other MS patients on the platform (with their explicit consent).

After the algorithm spits out a recommendation, it is passed on to the MS specialists at the university hospitals that care for the patients. These specialists both implement the suggested treatment and provide feedback to the algorithm. Treatment X worked for patient A, but didn’t work for patient B. It’s the perfect feedback loop to continuously improve the algorithm’s predictions and, with it, the care that can be offered to patients, a truly inspiring example of personalized medicine (which I touched on earlier, in chapter 6).

But there’s more I love about MiTrendS. It involves the entire (local) community in its mission. That includes patients with MS, of course. But it also includes healthy individuals who can use the app to help researchers establish reliable data for comparison.

You can’t understand a disease without tracking a patient’s symptoms. How do their neurological impairments progress over time? Are they able to complete certain cognitive tasks, and how well? But you also can’t understand a disease without having a clear sense of what to expect if those people weren’t suffering from MS. How well would regular people do at the task? How quickly do they get tired?

That’s what the MiTrendS citizen science part of the application does. It makes the village come together to support its most vulnerable members.

Making Data Co-ops a Viable Option

When I first heard of data co-ops a couple of years back, I immediately loved the concept. It sounded like a powerful approach to regaining all the fundamental rights we had lost in the transition to the digital economy. Privacy, transparency, self-determination. Data co-ops were designed to empower all of us; to not only take back control over our personal data and lives, but also benefit from the enormous value the digital economy had created (for a few big players rather than all of us).

As much as I love the concept of data co-ops, implementing them at scale is far from trivial. It requires us to fundamentally rethink the data ownership model and create an infrastructure that facilitates the collective management of personal data. But I am optimistic, mainly because we have pulled off similar stunts before.

When the Industrial Revolution concentrated power in the hands of a few major players, many citizens felt exploited and powerless. Over time, however, communities of individuals came together to form trade unions and citizen organizations that were guided by common interests and the desire to provide a counterweight to the big players.

Starting in the 1940s, for example, small member-owned electric cooperatives united under the umbrella of the National Rural Electric Cooperative Association to stand up against the energy giants of the time. Today, these cooperatives own over 40 percent of the electric infrastructure in the United States, covering more than 75 percent of the country. Not a bad outcome for people who started with no power at all.

Similarly, credit unions formed in response to the shift from traditional cash-based barter to digital consumer banking. When banks like J.P.Morgan threatened to dominate the market and exploit people for their own benefit, credit unions popped up all over the country. As nonprofit organizations with fiduciary responsibilities to their members, they started to offer many of the same financial services as traditional banks—minus the exploitation. Today, there are about 5,000 official credit unions in the United States servicing over 130 million individuals. That’s more than one in every three Americans.

The two examples show that shifting power back to the people is possible in principle. However, what makes me so optimistic about them is that they could lay the foundations of data co-ops.

As two MIT professors, Alex (Sandy) Pentland and Thomas Hardjono, have convincingly argued, credit or trade unions could be among the first and largest data co-ops.1 If you already entrust an entity to keep the lights on at home, negotiate your labor rights, or manage your investments and retirement fund, why not also entrust it with your personal data? It’s the simplest way to give millions of people access to a trustworthy advocate for their personal data, practically overnight. Or as Pentland and Hardjono phrased it in a joint report with multiples such unions, leveraging existing trade unions could make the “widespread deployment of data cooperative capabilities … surprisingly quick and easy.”

Regulatory environments—such as the European Union—which have already shifted data ownership to individuals by regulating data reuse and deletion, data interoperability, and portability, offer the ideal breeding ground for data co-ops. Your data is much more valuable when your data co-op is the only entity with access to it. As I mentioned in the previous chapter, bargaining becomes a lot harder if not only you have a copy of your data but everybody else does as well. It’s the combination of a competent crew and a relatively tame sea that enables you to make the most of your personal data.

Most importantly, however, a competent crew is valuable even when the sea is still rough. In fact, that’s perhaps when you need it the most—when sailing your boat alone is unlikely to end well.

Most US citizens don’t currently own their data. If you have ever tried to get access to some of the digital traces you generate, you’ll know how difficult (or even impossible) a task that is. In most parts of the United States, companies aren’t legally required to share your own personal data with you. At the same time, companies have the right to use, share, and sell your data to any paying third-party entity without notifying you (and these third parties, in turn, are allowed to do the same).2 Not a good spot to be in as a consumer. Left to your own devices, you have little to no power. Nobody is going to pick up your call and listen to your complaints and demands. But now imagine getting calls from millions of union members who are represented by expert lawyers. I bet someone is going to listen.