Authors:
Corentin Curchod – University of Edinburgh Business School
Gerardo Patriotta – Warwick Business School
Laurie Cohen – Nottingham University Business School
Nicolas Neysen – HEC Liège
Interviewer:
Alex Christian – University of Edinburgh Business School
Article link: https://doi.org/10.1177/0001839219867024
1. Hi Corentin, in your recent ASQ article, you and your co-authors illuminate the dark side of human-algorithm entanglement and how it can potentially elicit power asymmetries for those who work on online platforms. Can you please share how and when the idea for your article developed? Can you also elaborate on how your previous article on eBay from 2014 influenced your recent article?
The idea for “Working for an Algorithm: Power Asymmetries and Agency in Online Work Settings” initially emerged in 2010, when Nicolas Neysen, who was a PhD student at that time, and I had the idea to explore how online platforms impacted on the relationships between the two sides, for example between buyers and sellers on eBay. So, it all started from a general interest in online platforms and their intermediation roles. Interestingly, we initially collected data from eBay managers, not from the sellers of the platform. We didn’t use these data in the paper, but they helped identify our research object more precisely. In particular, eBay managers said something quite striking: they had problems with their business sellers, who were upset about eBay, and they did not really understand why.
Somehow, we felt that there was something going on here, which was interesting and worth studying more in depth. In this sense, our research was clearly phenomenon-driven. My co-authors and I are more inclined towards phenomenon-driven research designs rather than theory-driven ones. I also think that phenomenon-driven research is better suited for addressing practical problems. On the other side, the drawbacks of a phenomenon focus are that the theorizing process can be very time consuming and you have lots of data, so it is challenging to concentrate on a specific topic.
We started doing the interviews, and after a little while different topics emerged. One of the topics was about the identity work of digital workers and their difficulties to identify with the roles assigned to them by the platform owner. This is when Gerardo Patriotta joined the team, and we wrote together a paper that was published in Human Relations in 2014, “Categorization and identification: the identity work of business sellers on eBay”. The other topic that emerged during the interviews and that we explored further was about online evaluations and the impact on digital work. Laurie Cohen joined our team, and that was the ASQ paper.
2. Was it clear from the outset that you aimed to submit to ASQ? Why did you think your article was a good fit for ASQ?
For us, ASQ was a good fit because the article was theoretically ambitious, with our revisit of theories on power, and empirically convincing, with a strong and compelling story of contemporary work. But we had doubts, because ASQ has this mythical image of an inaccessible journal. So we asked friends and colleagues, and some of them said it was a risky target, hard to get in. We hesitated and had a few discussions together, where we would say “let’s go for ASQ” one minute and the minute after say “let’s go for another top journal”. Anyway, we finally decided that ASQ was potentially a good home for the paper and was worth a try. We thought that the paper would fit ASQ well, because the core-ideas appeared quite promising. ASQ likes a good story based on strong data, which changes the way one talks about a topic. I think the decision on where to submit a paper depends on the type of paper you have written.
3. How did you do the theorizing? What would you say is your main theoretical contribution?
A storyline emerged from the coding of the data, and this storyline remained the same throughout the revise-and-resubmit process. Yet the coding has evolved. It was an iterative process. At first, the storyline oriented us towards certain theories, on sensemaking and power. But the reviewers were not convinced by the sensemaking part. The first feedback from our editor Mike Pratt and the three reviewers was: “we love the story behind your data, but we think your theory does not quite fit the data, especially the sensemaking part”. As a result, we re-worked from the data, by going back to the first-order codes, and we realized that these fine-grained codes did not reveal much about sensemaking, but revealed a lot about our sellers’ experience of power. In the first version, we had force-fitted a theory into our data. So we agreed with the reviewers and decided to reorient the theoretical background of the paper around power-related dimensions. During the revision process, Gerardo, Laurie and I made continuous adjustments between theoretical contributions, second-order and third order coding. These elements evolved in parallel. It was a constant back and forth between theory and data. But the first-order coding, which stays very close to the data, has never changed. Metaphorically speaking, the bricks (first-order coding) remained the same, but the house (second-order and third-order codes) was built and consolidated progressively, with some changes made to the architect’s plan as well (theories).
By applying an inductive approach, we started by investigating the phenomenon and then interpreted it based on theories. In this way, the theory helped simplify the story and made it more impactful, understandable and useful for the reader, and reciprocally the findings contributed to the theory. In our article, if you take the theoretical story or the empirical story on their own, they are not as compelling as if you take them together. It is the combination of both that makes the arguments convincing. It’s an empirical and theoretical contribution.
For me, writing the contributions is the most difficult part of the work. The contributions evolved constantly during the writing process, as we always challenged ourselves, and the reviewers challenged us, asking to go more in-depth, to leave nothing in the dark. I think in the paper we contribute to an understanding of the link between evaluations and power in online platforms, and we show that the way people are assessed and monitored at work is changing because of the phenomenon of online evaluations. Compared to traditional work settings where you are evaluated by your manager, with online evaluations you are evaluated by hundreds of customers and by the algorithm, which compiles all these evaluations and triggers sanctions automatically. We capture these two levels of power asymmetry with the notion of “post-panoptic power” (Curchod et al., p.23), which is a nice analogy to understand the mechanisms at play on these platforms. But, again, it did not appear suddenly. It was an iterative process of writing and re-writing the contributions.
4- What role did your research design play in your research process?
Nicolas and I agreed on a research design from the very beginning. I remember that we discussed it together in a pub in Brussels. The research design was mainly about what the objective of our research was, and the best way to collect our data to achieve this objective. In our case, the objective was to understand the role of online platforms in the relationships between buyers and sellers. We thought the best way to start was to interview eBay managers, so they could explain how they saw their role in mediating buyer-seller relationships. As I said, when we met them, they explained they had a problem with business sellers, who were upset about the platform. We decided to refine our research design and focus on the sellers, to understand their work online, and their relationship with the platform. So the next questions were: how do we approach these sellers? How do we select them? EBay managers granted us access to online forums, where sellers expressed their discontent with the platform. We could have approached sellers via the forum to arrange interviews. But we thought that those who used these forums belonged to the same circle of very upset sellers, so we decided not to go that route. Instead, we defined criteria to select sellers and contacted them directly. This way, we avoided the ‘group effect’ bias and had a diverse population of informants.
Setting criteria was an important part of our research design. As we wanted to focus on online work, we targeted business sellers, for whom selling on eBay is a job. We chose sellers with enough evaluations to avoid inexperienced sellers, and we chose them among a mix of product categories to avoid interviewing for example only booksellers. To sum it up, the research design was how it all started. Even before we discussed our research methods, we set up a rigorous design, which was refined along the way.
5. Acknowledging that you published your article as a team of four authors, can you please elaborate a little on the specific roles each co-author played as well as on the experiences of working with co-authors? What are your recommendations to make this collaborative process as smooth as possible?
Nicolas Neysen was mainly involved in the data collection and the first-order coding, whereas Gerardo Patriotta and Laurie Cohen came when we started analysing and theorizing from the data. The collaboration with Gerardo followed up on our Human Relations’ collaboration. We decided to explore the power and evaluations aspects more in-depth, as this topic was very salient in the data. It turned out to be especially helpful to have Gerardo and Laurie looking at the case with fresh eyes and not being as involved in the data as I was. I took the role of the first author, which means that my job was to write successive drafts of the paper, based on data analysis and theory building. Gerardo took the role of the second author, which means that his role was to rewrite parts of the different drafts and build the theoretical contributions with the first author. Laurie had the role of the third author, which means that her role was to challenge Gerardo and I by proposing alternative interpretations and different ideas, helping clarify everything in the successive drafts. Being less involved in the writing of the paper, she could be our reviewer before the official reviewers, so to speak. So each author brought different skills to the paper, and the first 3 authors would meet at least once a month to work on successive drafts and discuss ideas. We turned out to be very complementary.
As for recommendations, I would suggest to be as clear as possible from the very beginning on who does what, to prevent misunderstandings. Be aware that lively and tense conversations are a natural part of the process. Therefore, it is important to choose co-authors you feel comfortable with. I personally think it helps if you organize the meetings in nice surroundings instead of a dull office. Location and atmosphere matter greatly, so you can for instance talk through your disagreements over a coffee!
6. What are the main impacts of your article for academia as well as for practitioners? And lastly, what would you recommend early career researchers aiming to publish a qualitative study in a well renowned journal?
I think because our paper starts from a phenomenon, it can have a great impact on both academia and practitioners. For academia, our article contributes to the discussion on algorithms and how they change the world of work. Our paper shows that working for algorithms offers freedom and autonomy to workers, but also disempowers them by putting them at the mercy of buyers’ unfair evaluations and depriving them of constructive communication to solve issues. It also shows that it is possible for them to become more resilient and able to cope with these challenges, in particular by ‘working around the algorithm’ and manipulating the data. This is because they use the platform every day, and develop a deep knowledge of the algorithm. They take advantage of some of its features and exploit the small gaps. In other words, they develop practices around the algorithm to cope with working for an algorithm.
For practitioners, I think it could make people aware that these online platforms change the nature of work and how you are monitored, assessed, rewarded and sanctioned. Algorithmic work can have an impact on the wellbeing of platform workers, because they may feel oppressed by something they don’t really understand or control. So, it could incite platform owners to change the way they design their evaluation system. For example, Uber has implemented a reciprocal evaluation system, differently from eBay’s unilateral system.
For early career colleagues, my first advice would be to pay much attention to the data. The quality of the data is paramount. I believe that qualitative research will struggle to be convincing if it relies on poor data or a shaky research design. If data are strong and rich, but the theoretical framework a little bit unsettled, it’s still fixable. I would compare this with cuisine: if you have bad ingredients (data), you can have the best recipe in the world (theoretical framing), it would still make an average dish. Whereas if you have really good quality ingredients, then even if you do not have a really good recipe, the dish will still be fine. If you also have both a good recipe and great ingredients, the dish will be amazing.
I have another advice: try to make the process enjoyable, because it can be a hard and long tiresome process. Reviews can be daunting even when they are positive. This is why you should find ways to celebrate the small victories. For us, each time we managed to go through a revision, we celebrated.
My last advice would be that places and work environments do matter. Personally, I sometimes need to change scenery and be in a nice coffeeshop, rather than in an office. I need to be in a place where I feel comfortable and inspired. I’m not sure this advice applies to everyone, but for me it’s important.
(Side note: we conducted this interview over a pint in a bar ;))
Alex’s Bio: Alex Christian is a Ph.D. candidate at the University of Edinburgh Business School, who is interested in the field of social informatics and thereby generally fascinated by how novel information technologies may impact society, as well as how societal needs may lead to changes in information technology over time. In specifically, his PhD project examines the social role of technologies deployed to address Covid-19. His research is mainly inspired by literature from Organization Studies, Innovation and Science & Technology Studies.
The ASQ Blog 