Traditionally, the study of algorithms has been seen as the subject of mathematics and computer science. Here, algorithms are dealt with from a utilitarian perspective: they are particular technologies to collect, sort and assemble information e.g. in search engines and encryption devices. They order and organize limited alternatives (Vormbusch 2012). Furthermore, algorithms not only sort information but also play games, calculate the probability of future events in risk assessment and weather forecasting, etc. As numerical expressions of problems to be solved by computational machines, they operate alongside strategic planning and thinking (Alt/Reischuk 2008). Thus, computer science defines algorithms as a combination of information collection and the solution of problem sets through the application of this information: "Algorithm = Logic + Control" (Kowalski 1979).
Furthermore, the translation of complex problems into the language of mathematics algorithms serves as a warrant for objectivity; irrational, emotional and erroneous behaviors are deemed excluded through this move. Often, the main difference between algorithmic and human behavior (routines, habits etc.) is seen precisely in this aspect of deviation, for an algorithm operates without feelings and with precision (Cormen 2013).
Very recently, the study of algorithms has migrated to the humanities, social sciences and cultural studies. While computer scientists, physicists and mathematicians usually refer to the utilitarian side of algorithms – how algorithms work, how can they be constructed – scholars in the field of culture and society have instead focused on the social and cultural consequences of applying algorithmic instruments to particular social problems.
One set of consequences are structural changes. Some scholars working within a Marxist framework argue that the use of algorithmic instruments in fact does not lead to structural change, since capitalist ideology gets inscribed in algorithms (Mager 2012) they are mere mechanisms in the reproduction of power (Granka 2010), the "Emperor’s new Codes" that serve "the enrichment of its leading participants", and are simply "rationalizing dominant distributional patterns" (Pasquale 2013). Others, however, have seen a change in the fabric of the social elite, a "take-over by nerds" (Ensmenger 2003): the application of complex technological devices and algorithmic instruments requires an elite with completely different skills, an elite predominantly trained in physics, mathematics and computer science. Historically, this is particularly apparent at the stock markets where the mathematization of the economy in the 1970s (Debreu 1991; Weintraub 2002) has led to a change in trading, shifting from a reliance on personal networks and interpersonal exchange to the mathematical methods of quantitative traders (Seyfert 2013).
Beyond these elites and structural changes, other scholars have pointed to the importance of algorithms as powerful carriers of meaning. Algorithms are performative as much as they symbolic, and may serve as "fundamental expressions of societies" (Sanz/ Stancík 2013). For instance, instead of simply showing how algorithms can be used to detect plagiarism in academic texts, Lucas Introna shows how their algorithmic evaluation also alters the definition of what it means to produce an 'original' text. If plagiarism is understood as (misre)presenting another's idea as one’s own, then algorithms cannot be said to detect plagiarism. Rather, they identify matching copies of chains of words. Accordingly, writers might adapt their style of writing so it conforms to the rules specified in the plagiarism algorithms. Plagiarism algorithms "detect the difference between skillful copiers and unskillful copiers", and thereby performatively produce the skillful copier as the "'original' author". The social consequence: an entire culture of selling 'original' seminar papers and offering ghost writing services emerges (Introna 2013). Hence, instead of treating algorithms as utilitarian devices, studies of society and culture identify the performative effects that accompany the algorithmic access to world.
Another consequence is a cultural and local specificity described as the "filter bubble"(Pariser 2011). Filter bubbles are created by the search engines and social media platforms that feed us with information that tends to confirm our opinions and political views. This tendency has been criticized for two reasons: first, these selection algorithms make it harder to gain access to outside information, creating a somewhat monological life, and second, these bubbles are created without our consent by internet conglomerates such as Google and Facebook. However, these 'bubbles' arguably provide what Anthony Giddens calls "ontological security", i.e., a trust in the continuity of the one's own culture (Giddens 1984). Online searches and newsfeeds might provide functional equivalents to the regularity and sequentiality of television programs that satisfy the viewer's needs for continuity and repetition (Silverstone 1983, Sanz / Stancík 2013). Sociologically speaking, filter bubbles might perhaps prove to be an element of cultural integration, providing a sense of familiarity in the potential infinite World Wide Web.
Beyond the perspective of performativity and integration, algorithms might also be analyzed from a narratological perspective. Thus, instead of simply asking if and how algorithms create coherent texts (Anderson 2012), thereby taking over the jobs of translators, writers and authors (Kushner 2013), a cultural analysis of algorithms might also look at the narratives contained in the codes themselves. This idea was expressed by Lorrain Daston who asked for "the history and mythology (in the sense of Roland Barthes) of the algorithm" (2004). This is a promising field that should be further developed. It offers scholars of culture and society the possibility of contributing new perspectives to the study of algorithms. An example may be seen in the algorithmic description and reproduction of creative processes such as writing and painting, which was the subject of a recent interdisciplinary exhibition at Konstanz University . By inventing painting algorithms those scholars attempt to make machines creative (Deussen 2013). However, Lorrain Daston's call also implies another perspective, namely, exploring the creativity that is already present in algorithms, even if they were not created with an aesthetic end in view. For instance, the narratives embedded in the codes of common algorithms, and their susceptibility to ‘mythologies’, have been explored by visual artists who transformed the narratives of various ‘sorting algorithms’ into images.
The conference Algorithmic Cultures aims at surveying the possibilities for a cultural analysis of algorithms in research and practice, as we have described above. We are interested in the social and cultural effects and consequences of the increasing diffusion of algorithms.
Questions and topics to be discussed in potential papers and/or panels:
1) How can the studies of culture and society develop the concept of the algorithm in a rich and complex way, beyond the tendencies in computer science, mathematics and informatics to focus on utilitarian aims? The question is first and foremost an epistemological one. For it is important to avoid using the term in an imprecise and inflationary way, by simply using it to refer to anything and everything that includes or involves software (see e.g. Bunz 2012). Do we have to understand the mathematical structure of algorithms in order to analyze their role in and effects on culture and society? This question of epistemology echoes concerns in other fields of theory and analysis that deal with highly “codified” or “ciphered” symbolic and cultural systems, such as Law and Economics.
2) Algorithms carry substantial implications for future developments in the field of socio-cultural theories. On the one hand, they appear to be universally applicable devices; insofar as they are applied in different fields, they seem to transgress social spheres and systems. For instance, a myriad of algorithms have been applied to stock market predictions, irrespective of the initial parameters and situations for which they were developed. These include algorithms for weighting game outcomes, for ballistic missile early warning systems (BMEWs), and for weather forecasting. Yet, each algorithm operates with an entirely different set of operational logics and expectations of future events. Such fluctuations seem to contradict the notion of a universal capitalist or economic logic, which penetrates each and every part of life world (see e.g. Pasquale 2013). On the other hand, algorithms are also not distinguishable according to clear cut social subsystems, each operating with its own rationality: algorithms with mutually contradicting logics can easily operate in the same system. It seems as if algorithms – as patterns of and logics for the solutions of problems –shift and move between and in different social fields.
3) Another sociopolitical issue arises around automation and human agency. Cultural Studies and its related disciplines abounds with accounts of how the increasing autonomy of algorithms leaves human actors as mere secondary conditions; this feeling of loss of control has led to calls to wrest back human control of algorithms. By contrast, computer scientists and coders of algorithms treat them as mere tools that can help enhance the productivity and dignity of autonomous human agents. For instance, financial traders have praised algorithmic trading for contributing to the common good by providing "liquidity" and increasing "efficiency". Both approaches operate with very strong notions of autonomy – only these are differently assigned, to human subjects and algorithms respectively. However, recent research in Social and Technology Studies and Social Studies of Finance argues that formulas, models and algorithms are not mere tools but rather an active component of social engines that shape the devices as much as the actors. As we pointed out before, algorithms might be conceptualized as performative – they are "engines not cameras" (MacKenzie 2008). Social and Technology Studies and Social Studies of Finance apply the conceptual framework of assemblages to their analysis of the operations of algorithms. The entanglement of algorithmic processes of decision making with preexisting interpersonal networks and relationships calls for a new conceptual approach to algorithms. Algorithms are not merely mathematical processes and computational devices but also contain certain social aspects, such as embeddedness in interpersonal networks. For instance, Beunza und Millo show how the New York Stock Exchange introduced algorithmic trading devices in 2005 by "folding" them into the previously existing social structures of specialists, financial intermediaries and experts (2013). Yet, even as Beunza und Millo attempts to overcome the naturalized divide between technology and society, the concept of "folding" problematically implies a universal logic through which either humans or algorithms control a particular outcome. Perhaps this issue can be addressed by conceiving of the relationships between algorithms and human actors as a "sociomaterial assemblage" (Introna 2011).
4) Still related to power and politics is the question of control, governmentality and institutions (Beer 2009). In Computer Science, algorithms are described as computational "recipes" (Press et al. 2007) – as sets of instructions for the (computational) solution of a problem. Similarly, Classical Sociology has described forms of normed and standardized human behavior that are used to solve recurring social problems. If we conceptualize algorithms as formal instructions for sociomaterial assemblages, then will these “recipes” differ, qualitatively or otherwise, from patterns of interaction that sociology has called imitation (Tarde), mimesis (Girard), habitus (Bourdieu), typified action (Schütz) and role play (Mead)?
5) The sociomaterial assemblages of algorithms create new types of unpredictability and thus unforeseen risks. Their assessment must therefore go beyond applied expertise. The numerous and unpredictable feedback effects of algorithms mean we cannot study algorithmic codes and their functions in isolation. The experimental tests of their behavior (e.g. in back-testing with historical data) cannot reveal the inter-algorithmic effects that exceed even the comprehension of those who design, manage and control them (Ullman 1997). What is more, the alterations introduced during the employment of algorithms (by software engineers) and their mutual feedback effects seem to reproduce the production of difference similar to that of human interaction chains. Not all problems can be expressed algorithmically—which is why those problems have to be translated and re-translated in algorithmically expressible language. It is precisely this "chain of translations" that makes algorithms and their interactions exponentially complex and increasingly "inscrutable".
First and foremost, by analyzing it from a cultural and social perspective the conference tries to approach the subject of algorithms from a different perspective than the one commonly found in the fields of Computer Science and Mathematics. While we are also concerned about the aspects of technical and political control (the main focus of the Governing Algorithms conference in May 2013 in New York) we are even more interested in the wide-ranging implications of their cultural diffusion – in the culture of algorithms. Such an approach will not only look at the integrative and disintegrative processes in the social structure that this technology entails e.g. the replacement or persistence of elites. It will also point to the meaning-making and the production of knowledge insofar as algorithms and their computational applications are not only mechanism to order and sort data but patterns to solve complex problems (a particular "knowledge logic", Gillespie 2012), patterns that move across different social and cultural fields and that entail new types of unpredictability and new forms of risks. Thus we are particular interested in contributions to Cultural Studies / Cultural Sociologies of Algorithms.