Second-order Data Exhaust

The footprints of data centers

tywen kelly
6 min readJan 11, 2021

“Pollution is nothing but resources we’re not harvesting. We allow them to disperse because we’ve been ignorant of their value.” Buckminster Fuller, I Seem To Be A Verb

Data exhaust are the footprints you leave behind as you go about your digital life.

Every action you take on a digital device is a footstep recorded in the sands of the metaverse. If you peruse the web, if you paint a canvas in Photoshop, if you open an app on your computer, you are generating many kinds of data exhaust: web cookies, image caches, and system logs. These files are the detritus of your being in a computer world. They are evidence of your virtual self having touched things with your mouse pointer; the alibi of having joined another realm. Collectively, these footprints carve a clear path of your behavior, tracing an outline of your shadow in between the negative space of your prints.

Data exhaust can be considered “first-order” if it becomes part of a feedback loop. A first-order feedback loop is defined through the study of cybernetics as a loop which is self-contained and closed off. If you looked at your own data exhaust — say your browser search history — and changed your behavior based off of what you saw, then that data exhaust has become part of a feedback loop. By observing your data exhaust, that data exhaust becomes first-order data exhaust, because you have simultaneously isolated the feedback loop and neatly closed it in on itself.

I doubt any of us really look at our data exhaust, let alone change our behavior after seeing it (although this is a good practice). Day to day we more realistically interface with another, prominent form of first-order data exhaust: our web cookies. Our web cookies are transformed into “first-order” status the moment an advertiser gets a hand on them and uses that data to give you something to respond to. An advertiser closes this loop and maintains the loop in the space between your behavior and the advertisers’. For instance, an online store might have cookies embedded in it, some of which follow you around between sites (à la Facebook’s Pixel), which tracks how you shop and allows advertisers to respond by adjusting or optimizes their shop to try and make you (re)buy their product or service next time. Scaled up, these first-order feedback loops between advertisers and a public can be called the “attention economy.”

But, you might say, the attention economy is not “first order” because it does not actually exist in a closed loop. It is a part of a greater economy, which itself is embedded in a nation-state, which is itself embedded in geopolitics, which is itself embedded in history, all of which are embedded in larger and larger loops. In fact, upon closer inspection, if a first-order loop is defined as a closed loop that touches nothing else and is in other words isolated, it begs the question if a first-order loop could ever even exist at all in the real world. Indeed, as data tends to connect and attract each other, almost like a force of nature, it seems improbable that any loop — let alone the loop of data exhaust — can exist without influence from extrinsic forces.

It appears to me that the metaloop to contain all loops is, on this Earth, Life itself. Life comes to us most definitely in the form of the infinitely complex system of nature. Our earth is in a feedback loop of its own with the sun, absorbing the suns energy and turning it into biotic material and a laying foundation for a natural landscape. The landscape is an emergent property of weather systems, nutrient cycles, and shifting tectonic plates looping into each other. Within the landscape are niches, habitats, and organisms each acting as simultaneous input and output for each other. As a gestalt sum, Life becomes the loop within all other loops inevitable fall into.

To have as existential context the metaloop of Life is important in framing the significance that the manufactured loops of something like an attention economy are having upon it. It implies that no loop is first-order. All loops connect to each other. This is what is meant by “second-order”: a loop that not only feeds back into itself but also is influenced by and influences other loops. The namesake of “second-order” comes from a new generation of cyberneticists, who, took the original definition of cybernetics, which is “the study of control and communication in the animal and the machine,” and looped it back upon itself. Second-order cybernetics is thus the “the cybernetics of cybernetics.”

Steam rises from vents at the Google Data Center in The Dalles, OR. The facility is situated alongside the Columbia River, which it uses to both power the plant through hydro power, and to cool the servers in the hot summer months. This is second-order data exhaust. (Image by Tywen Kelly)

Second-order data exhaust is data exhaust that has been passed through more than one loop. Pointedly, most second-order data exhaust is part of an extrinsic loop that is not yet closed. If first-order data exhaust is mostly ignored, second-order data exhaust is doubly so. Data centers owned by the like of Google, Microsoft, Apple, Facebook, and Amazon, are an example of a site of second-order data exhaust manifestation. The second-order data exhaust comes in the form of disspated heat, real estate infrastrucutre, electricity consumption, and the collateral byproduct of the towns and people that service these stations.

These facilities exist at the intersection of the physical and the digital. If you were to observer a data center you could see the heat shimmering out of its vents, evidence of an attention economy whirring behind its walls. That heat exhaust is the byproduct of deleting, overwriting, and reading of hard disks and cooling systems. It is a remnant of inefficient data cycles, an intentional design of computer chips that are optimized for clock speed rather than power efficiency.

The electricity that is consumed by these data centers themselves have a “carbon footprint”. The input (electricity) and output (heat, emissions, environmental infrastructure) are a part of the natural system, and are physically observable. The input comes from an unclosed loop, as the infrastructure required to power these plants have an imperfect, unbalanced relationship with the landscape as it stands. Hydro dams, for instance, change the landscape dramatically, affecting organisms like salmon, native peoples living nearby, and the very shape of the river as it carves through the rock. The electricity consumption of these facilities is huge: in 2018 data centers accounted for 1% of all global power usage, or about 205 terrawatt-hours. At data centers devoted to cryptocurrency mining, one building can take as much power as an entire country, in the case of Iceland. It has also been said that the carbon emissions are said to be on-par with the entire aviation industry, though recent efforts by large technology companies to invest in solar and wind energy has lowered the carbon emissions greatly.

The purpose of pointing out that second-order data exhaust exists as an environmental impact is to 1) show that digital bits and bytes have a material manifestation, and 2) that this material manifestation can be designed back into the loop. As the quote from Buckminster Fuller at the start of this article states, there really is not such thing as pollution, only material that is not re-introduced back into the feedback cycle. It shows that not only do people like us generate footprints, but that the things observing us generate footprints too, and that those footprints are worthy of re-capture into the loop. A hanging chad of data exhaust is nothing but harm to the landscape.

So, where does thinking about second-order data exhaust get us? I think it expands what it means for a digital metaverse to exist. It implies a mode of thinking that tells us that the metaverse is not separate from the greater loop of Life, that it is embedded and a part of it, and that it may have an unusually large (and exponentially growing) role in it. It means that when considering the bycatch of data mining and the debris of data storage we should have in mind that there exists a great-order loop, one in which is intrinsically and extrinsically connected to the one your are already in.