Algorithms and Rhetorical Agency

Recently, I’ve been thinking about the fitness monitors that more and more people are wearing to monitor their health. Of course, it's less accurate to say that people wear fitness trackers to monitor their health, than that they do so to monitor the monitoring of their health by the media technologies they’ve allowed access to their bodies. That difference matters. And it matters particularly because algorithms, on which these technologies run, are making it necessary to rethink the idea of agency.

Fitness monitors work first through sensors that measure input from our bodies, principally measurements of movement, and then through algorithms that interpret that data to generate an output feedback. It’s the output feedback that we monitor when we use these devices to track our health. So a Fitbit wristband or Apple Watch tells us our heart rate, or how many steps we’ve walked, or how fast we’ve run, or how soundly we slept, and so forth, and we respond (or not) to that feedback with the belief that it offers an insightful guide to help us toward a healthier life.

Insofar as the feedback offered in response to a fitness tracker’s measurements influences human action—e.g., actually prompts you out of bed to walk around the room until you log the day’s 10,000th step before sleeping—these devices can be understood to act rhetorically. If the algorithms driving fitness monitors can influence people, and if the process of exerting that influence can be understood as rhetorical, then in some ways these algorithms must exhibit a kind of rhetorical agency. The question is, what kind?

Carolyn Miller’s 2007 piece, “What Can Automation Tell us About Agency?,” is our best hope for an answer. Miller is the first rhetorician to recognize that automated technologies driven by algorithms challenge us to rethink what agency is and where it functions. For her, rather than “locate” rhetorical agency in a capacity to act rhetorically or in the effectivity of rhetorical action, as many have proposed we do, we should instead think of agency as a kind of kinetic energy attributed to the relationship between the two: between rhetor and audience, between capacity and effect.

Understanding rhetorical agency as kinetic energy might seem to "fit" fitness monitors perfectly, considering that these monitors inspire movement by measuring it. They're fundamentally kinetic. But if movement is both the necessary input for these algorithms to operate and that which the algorithmic output endeavors to inspire, then what agent function can be attributed to move-ability as such? In other words, movement needs to originate somewhere before the monitors and their algorithms can measure it, and hence before the algorithms can influence further movement. Is there not, then, a kind of rhetorical agency in potential energy as well?

Even if we follow Latour and attribute an agency function to an "actor network" (or, as I'm more inclined to do, follow Tim Ingold and call it a "creative entanglement"), the problem remains that we are not accounting for any originary sense of agency: for move-ability in the case of fitness trackers, but more generally for that ur-agency that sends agency's distributed capacities into motion to begin with. In a feedback system whereby human motion feeds a device’s sensors, which then feed output results back to the human, so that the human then moves or doesn’t move in response, thereby providing further input for the sensors, on and on, how can agency exist in any one node or in-between any of them?

An infinite loop has no start. Infinity has no middle. There is no in-between. Rhetorical agency, then, at least in this case, may well be the one that initiates or terminates the loop itself. In a world of algorithmic rhetoric, where algorithms play an increasingly prevalent role in mediating and influencing human affairs, that’s a powerful thought to remember.


One area where algorothmic rhetorical agency is expanding is into the health care industry. The data collected by your FitBit or Apple Watch can be standardized, stored in the cloud, integrated with other health data, shared with physicians, and used to monitor treatment efficacy in real time by patients and physicians alike. That infinite feedback loop picks up a few additional actors along the way — an additional cloud server and its connective and supporting networks; an algorithm that rewrites the collected data into a standard format for readability and portability; the larger health care network including its shared data across corporate and industry lines; and your physician and those inside and outside the office who support the office itself, including pharmaceutical reps and big pharma.

Take a look at Open mHealth, a start-up co-founded by Deborah Estrin, Professor of Computer Science at Cornell Tech. Open mHealth does everything I just mentioned with the goal of providing physicians and their patients real-time monitoring of health-related data for advanced diagnostic and palliative care. In a 2014 presentation, Estrin described the data collection and sharing functions of Open mHealth as a means of improving the timeliness and efficacy of treatment. Consider the following scenario: a patient visits a physician with a particular set of symptoms that the physician seeks to diagnose and treat. The treatment provided is a prescription medication which the patient is to take daily for the next two weeks; after two weeks, the patient is to return to the physician to evaluate the effectiveness of the treatment and determine whether the patient is symptom free — indicating the original diagnosis was accurate and the treatment appropriate to the symptoms — or retains those symptoms or has developed other symptoms — indicating the original diagnosis was inaccurate, in part or in whole, and requiring a new or revised treatment plan.

Open mHealth inserts itself by collecting and standardizing data that we’re already collecting with our FitBits, smartphones, watches, and sharing that data — with our explicit permission, of course — with the physician, who can use Open mHealth to poll and visualize the data to determine whether the treatment prescribed is, for lack of a more clinical term, working. In short, the physician who accesses our shared data can know within hours or days of prescribing a treatment whether that treatment is working. The physician can also uncover whether additional underlying causes or symptoms not immediately visible or shared during the office visit, like lack of sleep or sleep apnea or high blood pressure, might be reducing the efficacy of the treatment. The result, according to Estrin, is that more effective and proactive, palliative health care services can be provided using the data we’re already collecting, at a fraction of the cost of time, effort, and dollars expended toward attempting to treat illness without adequate or timely feedback from patient to physician.

Is this a good use of an algorithmic rhetorical feedback loop, where patient and physician communicate via algorithm-created and initiated data transfers? Where does agency reside when we expand the actor-networks or creative entanglements or rhizomatic systems to include so many more actors and agents?

This is a great example, Daniel, of the ways algorithmic technologies that are already part of our daily lives can be leveraged, if accessed, to serve different ends. In this case, as you mention, physicians using Open mHealth would need explicit permission to access your monitored health data, presumably checking all the necessary moral precaution boxes in doing so. But I'm suspicious of the outsourcing of surveillance to anyone, particularly to for-profit enterprises (even those ostensibly directed toward our health).

As for the feedback loop, I see the Open mHealth example less as a breach of that loop than as an illustration of the loop's recurrence within a larger actor network -- a sort of ripple in a pond, whereby we don't quite know what the pebble is. 

Chris, I was excited to read your piece because I have within the last couple of years taken up running and have just recently bought a Garmin Forerunner 225 GPS watch, using it to track my morning runs, my sleep patterns, my calorie burns, and more. It also functions as a pedometer and counts my daily steps. My husband bought one too at the same time I did and when he began noticing his steps (the watch automatically sets a goal for you and adjusts it up or down depending on your previous activity), he started saying things like, "Oh, do you want to go take a walk to [nearby store] after dinner? I need to make my steps today."

I found this compelling in the discussion of actors and agents you bring up above. Why did the watch compel him to feel like he needed to "make his steps"? Intriguingly, I routinely ignore mine--sometimes I "make my steps" above and beyond if I ran that morning or had lots of errands to do; other times I'm far below the number the watch suggests (those are writing days, usually), but I don't feel like I need to walk around the block just to make the watch congratulate me for hitting the step target goal.

So when you talk about "between rhetor and audience, between capacity and effect" here with regard to a fitness tracker, I wonder--who is the rhetor and who the audience? Who, exactly, does my husband worry will care if he does or does not make his steps? Why are some people compelled to walk further simply because their smart watch suggests they do so, while others ignore the suggestions?

Great questions, Stephanie. I certainly don't have definitive answers here. The rhetor and audience, I suppose, depends on which perspective one chooses to take. If you're a post-humanist, you'd have to be careful about framing the question without posing a false choice. If you're more traditional, it would depend what perspective you're taking: that of the algorithm/device (which, after all, "listens" to what your body "says"), or that of the wearer (who, of course, wears the device precisely because s/he intends to "listen" to what it "says"). As best I can tell, there's a mutually constitutive rhetoric here, almost a dialogue, such that both agents, device and wearer, are brought into rhetorical relation with one another in practice. 

As I read your post Chris, I immediately connected a rhetorical algorithmic agency with the Internet of Things movement to place sensor technologies within common household items. The Internet of Things movement is big business. International consulting firm, McKinsey & Company projects a 4 to 11 trillion dollar economic impact by 2025 due to the movement. 

With such an unfathomable impact, I think it's important to think about the rhetorical effects of algorithms in people's everyday lives. Just as your post illuminates, discussing agency in connection with algorithms will be an increasingly important topic to explore. At the same time, as I think about rhetorical effects I wonder about public education on algorithmic agency. How do rhetoricians teach the public about this concept (if at all)?

Isn't the Internet of Things just crazy? Of course, that's what I'd say in 2015. Ask me again in a few years. If the pundits are right, it won't be long before we can't remember or imagine an offline world of things. The challenge, I think, is to reclaim for our interdependence, not just among people, but within a broader ecology of things, plants, animals, etc., an appreciation for our deep relationality without relying on its networked, digital connections. This is an ethical move directed toward sustainability in a world where all our actions have a consequence for others, if by others we include indirect consequences and things that aren't just human. Algorithms, in a way, make these connections and consequences more visible, though of course not visible enough that their black-boxed rules are obvious to see. As pedagogy goes, then, my impulse would be to start with post-human moves to identify our offline interdependence within a broader ecology and then start looking at its more circuited (okay, wireless) connections in a digital one.

Add new comment

Log in or register to add a comment.