Paper short abstract:

Self-driving car assemblages are ordering devices. How does a self-driving car assemblage order its environment and what are the consequences of its ordering?

Paper long abstract:

Navigating a car in a large city can be confusing for inexperienced drivers. As for roboticists trying to make a car navigate itself, the messiness of urban traffic poses an equal if not greater challenge. Computers are great at solving well-defined problems, but how can you transform driving into a formalized problem? Self-driving car assemblages are ordering devices. How does a self-driving car assemblage order its environment and what are the consequences of its ordering?

There is little empirical research on the present versions of autonomous driving in urban traffic (Brown and Laurier 2017). My analysis is grounded in empirical material generated by original ethnographic research. Between June 2012 and November 2015 I conducted fieldwork among a collective of computer scientists and their computational enhanced cars in Germany. My research is mainly situated within conversations on robotics and artificial intelligence in STS (e.g. Collins 1995; Suchman 2008).

Paper short abstract:

GPS navigation routes are influential infrastructures that car drivers access via their SatNav devices. Based on my ethnographic fieldwork I follow the question, which logics are inscribed in these routes and how spaces are structured by computable quantities for this human/algorithmic routing.

Paper long abstract:

GPS navigation routes - fixed between an A and a B - are influential infrastructures though hidden behind windshields. On today's streets drivers individually construct and access these routes singularly for each journey via their SatNav devices. The drivers' navigational practices make the route visible by referring to the spatial contours of these digital infrastructures, which act both empowering and disciplining on their in situ decision-making along their paths. Navigating turns into a human-machine-interaction combining human skills of wayfinding with cascades of processes and calculations. By this, a heterogeneous set of actors meets along the route:

Programmers implement mathematical route finding strategies building upon the long-held standard of the "A star-algorithm". "Error tickets" are tranferred between automotive, software and map providing companies to optimize and repair map information for every specific street corner. At the same time drivers deliberate, whether the provided route from their device is reliable or not.

My paper presents negotiations from my ethnographic fieldwork and interviews with truckers, taxi drivers, messengers and software engineers. Based on empiric vignettes I reflect on new alliances that are created along this algorithmic infrastructure. I discuss the question which and who's logics are inscribed in these routes and how are spaces structured by computable quantities for this human/algorithmic routing.

Paper short abstract:

This paper examines how public sector organisations use social media management software as an interface for handling customer enquiries or complaints. It does this through a case study of Transport for London's use of social media management software to provide commuters with service updates.

Paper long abstract:

In the past decade, businesses and public institutions have increasingly made use of social media platforms to provide customer service to their consumers experiencing trouble. However, these organisations rarely use the social media platforms at the source - that is, they do not use the Twitter app or website to communicate to people. Rather, they use third-party social media management software that aggregates many social media accounts and content in one digital space. This, in effect, transforms social media platforms into an inbox for customer service staff to work their way through. But how have third-party social media management software become perceived as an essential part of the digital setting for demonstrating disruption? And how does social media data produced in customer service situations and through social media management software perform in the organisation well after the initial customer enquiry is completed?

In this paper, I explore these questions through an ethnographic study involving Transport for London's customer service team's process for demonstrating commuter disruption and responding to commuter enquiries. I describe their use of social media management software and how an initial customer tweet takes on a life within the organisation - for planning, performance review and, editorial decision making - for some time after the initial enquiry or demonstration of disruption. I also describe how consumer tweets provide justification to iterate the organisation's social media customer service provision, using a pseudo-Agile approach to question whether these practices are truly innovative or merely responsive to shifting specificities.

Paper short abstract:

This paper examines how the personalized local search engine Foursquare mediates its users' navigations and practices of place-finding in urban space, by profiling them as "subjects of locational tastes".

Paper long abstract:

By directing their users to potential destinations (restaurants, cafés and other places of interest around them), local search engines play an important role in contemporary practices of urban navigation and place-finding. On such spatial media, the definition of what is a relevant and valuable destination to display can be inscribed in very different ways in the app's algorithms. Are the most relevant places, the closest ones? The best-rated? Or the ones recommended by "experts"? Just like any map, these local search engines encapsulate specific meanings and values which they further display to their users.

This paper examines how relevance is inscribed in the algorithms of Foursquare, a local search engine, whose displayed destinations are personalized in accordance with each user's profile. In Foursquare's ranking of places, a certain understanding of the user as a "subject of locational tastes" is inscribed. The paper examines how this subject is framed as a bundle of past location data, momentarily connected to other data-producers (friends or calculated publics of similar spatial practices). Conceptually drawing on previous attempts to frame these data-subjects as "monads", "made up of complex, unique, dynamic and always varying metrics" (Ruppert 2012) and "distributed across varied populations of different kinds that intersect through them" (Mackenzie 2016), the paper asks: how does this ever-evolving, flexible and context-dependent mode of knowing subjects impact on users' practices of place-finding and experiences of urban space? From an empirical viewpoint, the paper builds on a study of Foursquare's uses in New York City.

Paper short abstract:

This paper argues that by uncovering the human-machine complexity of a navigation system, we can uncover the various layers of understanding, misunderstanding, and ignorance among engineers, their managers, and their machines, when building such systems.

Paper long abstract:

How is my car seeing the road? How is it processing what is happening around it? Who is also helping me drive? Today's car ride entails a multitude of computer-mediated maneuvers and routing procedures, with a car processing probe data such as road traffic and car speed. Behind the software that helps navigate the car, sit a team of software engineers who attempt to optimize a route - helping reduce the number of factors that can affect how a car gets from point A to B. Oftentimes, these engineers and other stakeholders involved in building such systems, understand only a fragment of its complexity. How do they hold it together? How do they avoid failure?

Based on 2 years of situated ethnography in a large mapping and navigation software company in Berlin, this paper zooms in on the multiple layers of understanding, misunderstanding, and ignorance in building a technical system. Where does 'knowing' a system begin and end? How much can we understand within the complexity of a computational system? In what moments do developers leave things up to chance? How do these barriers of understanding challenge hierarchical structures of power (such as power between a worker and a manager), or more experienced developer and a less experienced one?

This paper argues that in uncovering the human-machine complexity of a navigation system, we can reveal the various thresholds of understanding among engineers, their managers, and their machines.