Studies in Human-Algorithm Interaction
Case 1: Roboadvisors
Two papers have so far come out of an extensive ethnography of roboadvisors, lay financial technologies that use algorithms to offer optimized investment portfolios at very low cost and with minimal human involvement:
The active construction of passive investors: roboadvisors and algorithmic “low-finance” (2019)
(Forthcoming in Socio-economic Review)
How does algorithmic finance operate in society as it crosses the threshold of “low-finance;” that is, into the hands of lay investors? In this paper, I answer this question by examining a new phenomenon in the retail finance space: the roboadvisors, a class of digital financial advisor that provides advice and automated investment management online with minimal human intervention, at little to no cost…
I find that the algorithms of high-finance, such as high-frequency trading algorithms, have been re-cast to fit a market paradigm compatible with non-practitioners, i.e. passive indexed investment strategies that follow the prescripts of Modern Portfolio Theory (MPT). I also demonstrate that as relations are formed between algorithms of “low-finance” and lay investors, the end-users of these systems similarly undergo a process of reconfiguration. By actively constructing passive investors, roboadvisors and their algorithms discipline their users and objectify them through automating investment decisions and enforcing a principle of “don’t do” vis-à-vis the market.
Enacting the rational investor: non-professional market participants and the performativity of roboadvisors (2019)
(Under review at Economy and Society)
Non-professional (lay) investors appear to be the antithesis of rational economic man. They have been cast as less-informed, less-skilled, and less-knowledgeable than professional market practitioners, with an abundance of empirical evidence that they tend to lose money in the market, on average. This study builds the case that a new class of algorithmic financial advisor, commonly known as roboadvisors, enact non-professional investors as rational market actors…
This is achieved through algorithmic devotion to modern portfolio theory (MPT), which the roboadvisors embody, automate, and perform, conjuring Homo economicus into existence. Through this exposition, I further show roboadvisors to be a particular case of economic performativity where the model performs the actor, rather than actors performing the market through the widespread use of a model.
Case 2: Dairy Cow Monitoring Tech
I have been conducting an ethnographic study on dairy farms that use monitoring technologies on their herd, as well as interviewing commercial producers of these technological systems and academic dairy scientists. Field work is expected to finish in Spring 2019.
Ruminations of the Calculable Cow (2019-2020)
This project studies how new monitoring technologies and calculative devices employed in the dairy industry are reconfiguring generations-old practices and knowledges, and crafting novel socio-technological relations on the farm. The cow becomes re-understood as screen-based data, changing the practice of the farmer and rationalizing the messy business of dairy farming. ..
Calculative devices, equipped to cows, transform the organic animal into a consistent and auditable stream of data, where the biological and behavioral facts about individual cows and aggregated herds are translated and projected onto computer screens and mobile devices. Data points and deviations from herd averages or a cow’s own individual data history indicate the optimal time to impregnate, predict when an animal is unwell, and increases milk productivity. By relying on data instead of the tacit knowledge and visceral, direct physical interactions with cows, the algorithm, cow, and farmer enter into a new sociotechnical arrangement defined by market logics of rational action. While this alienates the individual from the animals, at the same it time gives each individual cow its own “voice” that the farmer can acknowledge and respond to. The rumination of the cow – i.e. behavior around the chewing of cud – is constructed as an index of overall well-being, as a predictive marker, and as an irreducible re-interpretation of the biological cow around which practice is ordered. Before these technologies, rumination was known to be an important heuristic for bovine well-being, but it was never achievable as a numeric value, let a alone a value that is standardized between animals, across farms and farmers, and comparable over time. Tacit knowledge becomes explicit and creates new ontologies where the cow exists (is enacted) in a plurality of states simultaneously: both biological and messy, and technological and rationalized. Rather than the farmer understanding the factual cow differently as he shifts his worldview, the calculable cow exists in a quantum state constituted by its multiple forms.
Questions of organizational agency also arise. As the farmer comes to trust and rely on the data-version of the cow more and more, he risks being punished by the market if he deviates from the algorithmic decision-making (when to breed, when to treat an ill cow, when to cull, etc.). Indeed, if the algorithm instructs the farmer to breed a particular cow and he ignores that command, he risks an economic penalty when the cow fails to get pregnant using traditional practices or heuristic methods of timing estrus. Ultimately, the farmers I’ve interviewed stop second-guessing the system and submit to its instructions. Is the farmer managing the algorithm or is the algorithm managing the farmer? Since the data fed into these algorithms are ultimately bio-sensed from the cows’ physiology and behavior, is the farmer managing the cows or are the cows managing the farmer? What happens if and when the technological system fails – are generations-old knowledges destroyed? Do new knowledges about farm practice arise as algorithms train their farmers? What are the implications for both manual and managerial labor when algorithmic decision-making takes precedent? While the pastoral image of a dairy farm presents a striking and unique case of socio-algorithmic assemblage, these questions are indeed generalizable off of the farm as algorithms pervade every niche of economy and society from Wall Street to Main Street to rural pastures.
Case 3: Blockchains & Digital Currencies
In addition to publishing papers on cryptocurrency pricing models, I am also interested in the study of Blockchains and decentralized cryptocurrencies from a sociological and social studies of finance perspective.
The Socio-Technological Lives of Bitcoin (2019)
(in Theory, Culture and Society, 36(4), 49-72)
Through a historical sociology of the antecedents and discourse leading up to Bitcoin, I show that it was never meant to be “money” in the economic sense, but rather a solution to a technical puzzle for preventing opportunistic actors from double-spending digital “coins,” as well as a fervent ideology surrounding online privacy and infringement of individual rights in the digital age…
Bitcoin, cryptocurrencies, and blockchains have become buzz-words in the media and are attracting increasing academic interest, mainly from the fields of computer science and financial economics. In this essay, I argue that cryptocurrencies and blockchains are important objects of general social science research and thought, but not for their “moneyness” per se. Drawing from themes in science and technology studies, I suggest that Bitcoin and other “cryptoassets” are properly socio-technological assemblages that constitute new and important objects of social inquiry that must be understood beyond the myopic context of crypto-money. I conclude by proposing three alternative ontologies for blockchains relevant to economic, political, and social life: as systems of accounting, as organizational forms, and as institutions in their own right.
The evolution of the bitcoin economy: extracting and analyzing the network of payment relationships (2018)
(With Paolo Tasca and Shaowen Liu, in The Journal of Risk Finance, 19(2), 94-126)
This paper aims to gather together the minimum units of users’ identity in the Bitcoin network (i.e. the individual Bitcoin addresses) and group them into representations of business entities, what we call “super clusters”…
While these clusters can remain largely anonymous, the authors are able to ascribe many of them to particular business categories by analyzing some of their specific transaction patterns (TPs), as observed during the period from 2009 to 2015. The authors are then able to extract and create a map of the network of payment relationships among them, and analyze transaction behavior found in each business category. They conclude by identifying three marked regimes that have evolved as the Bitcoin economy has grown and matured: from an early prototype stage; to a second growth stage populated in large part with “sin” enterprise (i.e. gambling, black markets); to a third stage marked by a sharp progression away from “sin” and toward legitimate enterprises.
Decentralized Banking: Monetary Technocracy in the Digital Age (2016)
(In Banking Beyond Banks and Money (pp. 121-131), Springer)
Bitcoin has ushered in the age of blockchain-based digital currency systems. Secured by cryptography and computing power, and distributed across a decentralized network of anonymous nodes, these novel systems could potentially disrupt the way that monetary policy is administered – moving away from today’s human-fallible central bankers and towards a technocratic, rules-based algorithmic approach…
It can be argued that modern central banks have failed to stem macro-economic crises, and may have, in fact, exacerbated negative outcomes by incentivizing excessive risktaking and moral hazard via unconventional monetary tools such as quantitative easing and negative interest rates. A central bank typically serves three primary functions: to issue and regulate the supply of money; to serve as clearinghouse for settlement of payments transactions; and to serve as lender of last resort. Could a digital currency system serve as a rational substitute for a central bank? This perspective paper examines that question, and then suggests that indeed it could be plausible. While Bitcoin in its current form will prove to be inadequate to function as monetary authority, I put forward what an operative case could resemble.
Cryptocurrency value formation: An empirical study leading to a cost of production model for valuing bitcoin (2016)
(In Telematics and Informatics, 34(7), 1308-1321)
This paper aims to identify the likely determinants for cryptocurrency value formation, including for that of bitcoin and develops a cost of production pricing model. This suggests that bitcoin is more like a produced commodity than a financial asset or currency…
Due to Bitcoin’s growing popular appeal and merchant acceptance, it has become increasingly important to try to understand the factors that influence its value formation. Presently, the value of all bitcoins in existence represent approximately $7 billion, and more than $60 million of notional value changes hands each day. Having grown rapidly over the past few years, there is now a developing but vibrant marketplace for bitcoin, and a recognition of digital currencies as an emerging asset class. Not only is there a listed and over-the-counter market for bitcoin and other digital currencies, but also an emergent derivatives market. As such, the ability to value bitcoin and related cryptocurrencies is becoming critical to its establishment as a legitimate financial asset.
Bitcoin price and its marginal cost of production: support for a fundamental value (2018)
(In Applied Economics Letters, 26 (7), 554-560)
This study back-tests a marginal cost of production model proposed to value the digital currency Bitcoin. Results from both conventional regression and vector autoregression (VAR) models show that the marginal cost of production plays an important role in explaining Bitcoin prices, challenging recent allegations that Bitcoins are essentially worthless. Even with markets pricing Bitcoin in the thousands of dollars each, the valuation model seems robust. The data show that a price bubble that began in the Fall of 2017 resolved itself in early 2018, converging with the marginal cost model. This suggests that while bubbles may appear in the Bitcoin market, prices will tend to this bound and not collapse to zero.