As a technical mediation, Gephi shifts your goals (and vice-versa)

5 min. read

We had to cut this text from a paper we wrote with Emilija Jokubauskaitė, but I find it useful, so I share it here, slightly reworked. We borrow Bruno Latour’s (1994) explanation of the mediation, and provide examples about Gephi. Latour’s approach draws on Gilbert Simondon’s thinking of the technical object (1958), and more recently on Madeleine Akrich (1993). It contends the existence of a script, a program, attached to the object; and that program is neither determined by just the practices, nor just the instrument. It helps us make a very important point: Gephi influences its users by shifting their goals.

Have you been tempted to share a Gephi screenshot even though other people would not understand it? Gephi’s image-production abilities can shift its users’ goals.

Latour (1994) observes that there is a “materialistic” and a “sociological” version of the question about whether (and how) tools influence us. In a “materialistic” perspective, “each artifact has its script, its ‘affordance’.” This version emphasizes the role of material-semiotic features over practices. And there is a “sociological” perspective where the tool is a “neutral carrier of will that adds nothing to the action.” This version emphasizes practices and rhetorics, and puts the blame on people. As Latour notes, “the two positions are absurdly contradictory.” Analyzing technical mediations requires getting out of these false alternatives. When it comes to the criticism of data visualization, I find for instance that Johanna Drucker’s statement that “graphical tools are a kind of intellectual Trojan horse” (2011, 2014) sounds absurdly materialistic (more quotes in this post). Latour offers four meanings for mediation: “translation”, “composition”, “reversible blackboxing” and “delegation.”

Translation refers to the ability of the user-tool assemblage to displace the goals of both the user and the tool. The goals of researchers using Gephi may differ both from the goals of unequipped researchers and, symmetrically, from the goals of Gephi not in use. Producing images that can circulate may arise without being a goal of Gephi. “Responsibility for action must be shared among the various actants” (Latour, 1994). Let me explain. Gephi has been thought mainly as an exploration device. Even the image-export features were aimed at exploration: their purpose was to print images to annotate and discuss networks on a paper support, which is often better than the screen. We have stated it multiple times (Bastian et al., 2009; Jacomy et al., 2014). Yet I have seen countless Gephi screenshots on Twitter or in people’s slides that are shared to an audience who cannot understand what they mean, because they lack context and/or graphic design. Yet these images produce an effect. For instance, they showcase technical mastery, or the complexity or massiveness of the data. And that narrative might even be unintentional, which I call storyletting: letting your visualizations tell their own story, because you share them without a narrative. Producing such images is not Gephi’s goal, and I assume that it is not the researchers’ goal either. These images are brought into existence by the exploration process, and are legitimately stored as traces of the process. The same way Lada Adamic and Natalie Glance (2005) featured the famous figure 1 of Divided they blog to document their process. These screenshots are not intended as marketing assets, but they may become ones by circulating. This shift in their significance is what Latour calls “translation.”

Composition refers to the fact that the tool enables certain things. Even if we attribute agency to the researcher and not to Gephi, what the equipped researcher is able to do is made possible by both what the researcher can do, and what Gephi can do. Like translation, this meaning of mediation emphasizes symmetry between the tool and its user. For instance, Gephi allows seeing clusters, which is not visible in the raw list of nodes and edges; but of course Gephi symmetrically needs you to see the clusters.

Reversible blackboxing refers to the fact that the tool’s transparency is relative and ever-changing. When a device breaks, it has to be unblackboxed to be repaired, and can get blackboxed again to be put in use. The archetypal example comes from Merleau-Ponty’s (1962) phenomenology of perception: the blind man’s stick. When in use, that tool is embodied. The blind man ceases to feel it in their hand. It becomes part of their body, and they feel through it. Similarly, your glasses work insofar as you do not see them; if you see them, then they are dirty and that is a problem. Of course you need to look at your glasses to clean them, then you remove them, you disembody them, you open the black box for maintenance. And when you put them back, you return to blackboxing. Similarly, a researcher may consider Gephi a black box at the moment of using it, as usage requires a form of convergence between the tool and the user. The same researcher may open the black box at a later moment, when it comes to write down the method in a publication. There is a moment to discuss the algorithms you use, and there is a different moment when you want the algorithms to work. Researchers navigate between these moments. Gephi may be more or less blackboxed in different situations, at different moments, or to different people.

Delegation, the “most important” meaning of mediation according to Latour (1994), and based on the previous three, refers to the ability to shift signification. An apparatus may play a rhetorical role, it also has a meaning that is part of no discourse. Technical objects “act, displace goals, and contribute to their redefinition.” By shifting goals, the technical mediation ultimately shifts meanings. I find it a good explanation of why Gephi is sometimes understood as a machinery to produce images. Of course visualization is its purpose; but that is intended for exploration: visualize for yourself. Yet Gephi has somehow become a way to produce images for other people. Instead of wondering whether to blame Gephi’s program of action or user community, which would be returning to the materialistic and sociological versions of technical mediation, we can now understand that the [researcher+Gephi] composite’s new goals have shifted the meaning of Gephi towards the production of circulable images, if not marketing assets. This does not mean that one cannot intervene onto Gephi’s design or community, but it reminds us that something intended by neither its designers nor its users can nevertheless emerge from the interactions with the instrument.


Adamic, L. A., and Glance, N. (2005) The political blogosphere and the 2004 US election: divided they blog, Proceedings of the 3rd international workshop on Link discovery, pp. 36-43.

Akrich, M. (1993) Les formes de la médiation technique, Réseaux, 60, pp. 87-98.

Bastian, M., Heymann, S. and Jacomy, M. (2009) Gephi: an open source software for exploring and manipulating networks, Third international AAAI conference on weblogs and social media.

Drucker, J. (2011) Humanities approaches to graphical display, Digital Humanities Quarterly, 5(1), pp. 1-21.

Drucker, J. (2014) Graphesis: Visual Forms of Knowledge Production (Cambridge, MA: Harvard University Press).

Latour, B. (1994) On Technical Mediation, Common Knowledge, 3(2), pp. 29-64.

Jacomy, M., Venturini, T., Heymann, S., & Bastian, M. (2014). ForceAtlas2, a continuous graph layout algorithm for handy network visualization designed for the Gephi software. PloS one, 9(6).

Merleau-Ponty, M. (1962) [1945]. Phenomenology of Perception. transl. Colin Smith (New
York: Humanities Press).

Simondon, G. (1958) Du mode d’existence des objets techniques (Paris: Aubier).

Cite this blog post
Mathieu Jacomy (2021, March 16). As a technical mediation, Gephi shifts your goals (and vice-versa). Reticular. Retrieved June 18, 2024, from

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Search OpenEdition Search

You will be redirected to OpenEdition Search