Poetry and Plumbing in the Internet of Things
This is a talk I gave in Fall 2016, at the Marketing Sciences Institute
Three years ago, I left my job as a sociology professor at Barnard College, where I had worked for almost a decade, and moved across the country, from New York to Portland, Oregon. I took a position as a “senior research scientist” at Intel Labs, a research and incubation organization within Intel. I am almost certainly the only sociology PhD in a company of 100,000 people. I went there to work with a handful of other social scientists, technologists, and designers to learn about, and to shape, the future of technology – that is, to do foundational product innovation. We did this work by going out and researching people in the world, building prototypes, and designing systems that we thought would intercept people’s changing lives.
Three years later, I now work as a “strategic planner” in Intel’s Datacenter Group. My work today is focused more at the intersection of social science, technology, and strategy. I help identify competitive threats, transformational ecosystem changes, and basic business model innovation. At its heart, what I do at Intel is provide context, frameworks, and tools to determine how ecosystems are changing, and what those changes mean for the company.
As part of this, I have conducted research on precision agriculture, helped design a system for distributed, next-generation transactions, and partnered with external companies to build and test all kinds of prototyped technologies.
I am sharing with you today, a view developed with colleagues, of the future of IoT. My framing of this view as “poetry” and “plumbing” borrows from the organizational sociologist Jim March. His argument is that leadership is a combination of poetry and plumbing. That poetry renders meaning into action: those “interesting interpretations of reality” that form the basis for “constructive collective action.” On the other hand, no organization works if the toilets don’t work. Poetry provides vision. Plumbing makes that vision work reliably and routinely.
In the tech and marketing world, most people appreciate the poetry, that vision of the future capturing the imagination and bringing the impossible into focus. Self-driving cars. Smart Homes. Actuating technology, reaching across the human-digital divide, enabled and personalized through data. And unsurprisingly, most people are, at best indifferent to plumbing: the nitty gritty user research, technological development, coding, integration, and ecosystem work that may - just - make the future possible. Absent the plumbing, we get the non-functional and trivial: a world of crappy robots and self-ordering refrigerators.
First, then, is some poetry around the consumer Internet of Things. That is, the vision of what is possible when we migrate from computers on desks in offices and homes, toward compute in the external environment. And then some plumbing – the work of shaping the data collection and analytics that it would take to make that poetry sing.
We are living in a signal moment, a time of great technological transformation. And yet, the shape, process, and dispositions of this transformation are still fuzzy, at best. AI, unprecedented compute power, and social networks now effortlessly connect billions of people and organizations around the world. But we also face unprecedented concentrations of power and wealth, dramatic changes in meaningful work, global environmental climate change. We are, as my colleagues would say, in a state of “Capital-F” Flux.
For example, we have all likely experienced the transformative effects of mobile platforms. “On-demand” (for those who can remember checking and printing out customized maps before making a trip, like animals) is now something entirely new. Waze, Snapchat, Uber, Pokémon Go, Fitbit, Amazon Echo. These apps, services, wearables, and devices don’t even really live natively on your computers anymore. I live in Portland, where you need a phone to use the Bike Share. Whole economies are now built on mobile platforms that simply did not exist just 10 years ago.
However, at the same time, embedded systems – machines operating in tight coordination in factories, navigation and flight systems operating in tight coordination in jets, which are the precursors to IoT – are at this point half a century old. “Compute in the environment” is both old, and new.
We need ways to distinguish between the magic and mundanity of this moment. It’s not “things connected to computers” that’s new. What’s new is things connected to computers, to networks of other things, and to our natural environment. And of course, these things connected to us. We are just starting to understand the transformations happening around IoT. I’m going to share two of them.
The first insight is that IoT worlds are “model-to-measure” worlds. Cheap sensors, analytics, and platform tools change our expectations and challenge our assumptions around long-held knowledge and practices based on “models.” Models are meant only ever-so-slightly metaphorically. They include true models: actuarial tables used by insurance to value your car insurance, semi-annually, based on the make and model of your car, your age, and your reported driving history. They also include estimates based on historical averages and folk knowledge: that bees are monogamous; that the corn would be knee-high by the fourth of July; that you should exercise 30 minutes, 3 times a week; that you milk the cows twice a day. I could spin this out even wider, still: Yelp reviews as proxies for judging restaurant service; 400MG of Ibuprofin, based on the average weight of a typical US adult.
What changes with IoT is, it turns out, nearly everything. From “model,” where your car insurance is based on your demographics, calculated every 6 months, to “measure,” where pricing is based on individual behavioral data, based on the last 300 miles you drove, calculated daily. A “measure” world is one where farmers feeding corn to cows based on the number of calories that cow ate in in the pasture, 30 minutes ago. A measure world includes: restaurant service measured as time-to-table, via sensors embedded in the serving trays; the “green-ness” of Intel’s supply chain documented with verifiable measures; and, individualized, precision medicine based on your personal DNA and daily well-being measures, measured in near real-time.
Many of those visions remain just that: visions. But this world is already technologically possible. The future is already here, it’s just unevenly distributed. And what we see is that insurance is getting there first. Over the past two years, Progressive insurance has begun, and increased, a program they call Snapshot. It is part of a wider move toward UBI, or usage-based insurance. Using a telematics device, they capture real-time data (miles, hard brake events, and what time you drive), to price insurance not based on who you are, but based on what you do.
From auto insurance, this model-to-measure frame is migrating to home insurance, health insurance, toward more esoteric forms of risk management. It is very much in the process of disrupting actuarial professionals, who are experts in post-event financial effects analysis, but who know almost nothing about real-time, pre-event, condition and hazard level monitoring.
However, moving towards measurement is necessary but not sufficient to unleash durable value. This leads me to my second bit of poetry: that durable, platform-based value in IoT requires us to understand data’s circulatory value. This is an analytical distinction between “primary” and “circulatory” uses of data. Primary data is gathered and analyzed in order to serve a particular purpose; and that same data as it is combined, analyzed, and repurposed in alternative contexts is circulatory data.
Primary data is you, using your health tracker to keep tabs on your daily steps. Circulatory data is the use of your data with others, then used by healthcare providers to determine best practices; insurance providers using that data to price your health insurance; and government agencies using that same data to understand public health. Farmer’s want to measure how much water and fertilizer they need to put on their fields; banks are interested in using that same data to price the costs of capital.
This is, of course, a somewhat psychotic view of how your personal data gets used, traded, mobilized, and then, absent regulatory structures and existing mores of privacy, used against you. But it hopefully illustrates what I want to convey: that the circulatory value (exchange value and use value) of your personal data is almost certainly greater than that data’s primary value.
It is also a slightly psychotic view in the sense that almost no data today circulates thoughtfully. Or rather, the ‘circulation’ is a stunted, highly-impoverished circulation between you and the advertising industry. Much of current-day IoT is one-off, everyone for themselves. Your lights talk to your alarm clock; perhaps more suggestively, your car talks to your mechanic.
More frustratingly, in our research, we have found that no line of sight is the case even in areas where there could be high value in circulating data – at Portland airport the parking systems don’t speak to the concession systems don’t speak to the air traffic logistics don’t speak to the WiFi providers.
To the extent that there are suites of IoT- smart cities is perhaps the best example – what we get is the hope and dream of long-tail value discovery: a bit of traffic, some pollution monitoring, water systems, and social media use. One company we looked at was trying to track tourists, to find out where they go and why, to “identify emerging market trends.” What trends? The data will tell us!
Now, we have a running joke in Portlandia around “farm-to-table” practices. Is it local? And “local” today is measured by certifications – USDA Organic. It is possible today to measure the environmental impact of a dairy cow in its pasture (how much effluence and methane gas it releases), trace that animal as it moves through the slaughterhouse (how many days it waited, how many people touched the animal), onto a truck (how many miles it drove, and at what temperature), and into the market (where packaging, measuring bacterial growth in real-time, would turn color when the meat is no longer fresh). Goodbye, and frankly good riddance, to expiration dates.
So, that joke is no joke – it is a form of IoT synergy along a highly integrated value chain, providing massive amounts of value all along the way. You are simply most likely to find circulatory value in IoT data, in areas that form natural, integrated value chains.
But here is the rub, the unsexy toiletry I promised at the outside. If you want to move towards a world of model-to-measure, and you want your data to circulate with value, it is fundamental to be thoughtful about how that data is captured, identified, and exchanged. This is what I mean by data provenance.
We have found fairly consistently the two main barriers to data circulation are contentiousness over ownership and contentiousness over value: who owns the data, who derives value from the data, and who gets punished by the data. This extends to models and algorithms. If I withdraw my data, does the data disappear? Does your model get refreshed?
The key transition here is from ‘sharing’ to ‘exchange.’ A working system must allow people to define boundaries within which circulation is allowed. There is big difference between transparency around what sensors are actually measuring (resistance, capacitance, and induction), meaningful data (and attendant metadata), and algorithms. Without a regime of exchange, we are too often left not knowing what we are giving up, for what we are getting.
In similar fashion, circulatory data only has value when primary data can successfully travel – what sociologists would call commensuration. Primary users have high levels of specific knowledge about their fields, houses, machines, processes. You wear your fitbit on days you exercise; you know your Nest runs 3 degrees too cold because it sits in your drafty hallway.
For additional users to get value from IoT data, those users need sensors to be placed appropriately. Without strict compliance, additional users need metadata about the placements (“this sensor is in the wettest part of the field,” or “where the sun shines most”). Without metadata or strict protocol compliance, the data are only useful to the primary user.
With good commensurability (good sampling, or great metadata), something magical happens. Sampling and circulation begins to challenge our ideas of community – who makes it up, and how much proximity matters. Just as usage-based insurance reveals that I drive just like a 20-year-old woman from Austin, as well as a 78-year-old man in Bethesda, so too does it change who my physical community is.
If I had more time, I would tell you more. I would wax poetic around the massive increases in connectivity that are around the corner, the decentralized machine-to-machine transactions that are on the cusp of being enabled. And I would dig into the compute topography that is required to efficiently allocate data at the edges of IoT networks, alongside the business model development required to make those changes feasible.
But in the meantime: consider migrating to measurement, along chains of circulating data; practice good data stewardship alongside rigorous collection protocols. The opportunities are there if you are willing to embrace them.