Angela —
Your recent book is called Autographic Design. What is autographic design?

Dietmar —
In the book, I try to call attention to material expressions that can be compared to visualizations, but are non-representational, non-symbolic. One starting point for me was looking at practices of citizen scientists or citizen science practitioners or environmental activists, or people who engage in amateur forensics. Despite there being a huge push to educate these groups in traditional data visualization approaches, they would typically use analogue methods to share original data material. I think it is because the rise of disinformation has caused the public to be quite critical of data that is presented in a traditional scientific manner. People want to know where data comes from and how it was analysed. If this is not explicit in the visualisation, it invites scrutiny.

Public discourse and public controversies are not about whether data are correctly interpreted, but about how data were collected and analysed, and in which context. When non-data experts focus on the metadata materiality of the physical source material (the artefact), they can make claims that are difficult to dispute. This is very different approach to presenting information and making claims about facts and evidence.

Angela —
Can you provide an example of self-inscribing data?

Dietmar —
Sure, the New York taxicab dataset contains millions of GPS points locating where New Yorkers get picked up and dropped off in taxis. When you plot these locations as points on a map you see some very blurry areas and some very sharp areas. The blurry areas are a result of the GPS signal being blocked by tall buildings, because GPS signals are less precise when tall buildings are nearby. So, when you plot the points, they fall in ‘clouds’ near taller buildings because the accuracy of the location of these points is more “noisy”. The blurriness of the data-point clusters corresponds to the height of the buildings in that area. So in order to infer a three-dimensional condition, you only need two variables (latitude and longitude). These two variables can encode the three-dimensional shape of the city. As a data designer, if you look at the blurry areas as a data artefact, instead of as a mistake that needs to be eliminated, a third hidden variable is revealed to you.

Color plate by Francisco Javier de Balmis shows smallpox vaccination scars. Courtesy of Welcome Library (published in: Orkan Telhan and Dietmar Offenhuber. The Inscriptome: Virus as a Visual Instrument. Strelka Magazine. 08.09, The Revenge of the Real. September 2020)

Angela —
Many data researchers would try to remove that blurriness because the location accuracy is not precise, and they might regard these points as error, instead of seeing what it reveals about the city.

Dietmar —
Every digital dataset has these hidden self-inscriptive aspects. We usually think of data as something representational, where a symbol refers directly to a real phenomena in the world, but I am not interested in the symbols of data, but in physical traces of the world that are self-inscribing, in the traces that do not represent or symbolise anything other than themselves.

Working in an autographic way is also about process. The process admits that when you don't know the causes of these hidden dimensions or traces, you have to speculate. Part of data analysis includes the personal stories we tell ourselves while trying to interpret data. Myself, I work in a very speculative and counterfactual way to consider the imaginary possibilities of data. Data seen in this way is no longer a bureaucratic or fact-based artefact that instantly tells me what to do or what decision to make. It is a speculative process. And this is a powerful acknowledgement for political or decision-making processes where we don't admit that, often, data is included after the fact as a justification for decisions that were already made.

Angela —
So, you are like a data ethnographer. You are a part of the phenomenon, and you also leave traces.

Dietmar —
Exactly, yes. There is also a certain playfulness. I think that has always been a central part of both data visualisation and data analysis. But it's not always presented that way. It's not always admitted.

Angela —
Can you talk a bit about the process of autographic design, what does it look like?

Dietmar —
From my perspective, it's about being very deliberate. There is a lot that happens between receiving a dataset and representing it on a screen. You need to aggregate, you need to reformat… and you should be very intentional about that. We open or reveal a process that is normally hidden, rather than just showing the result. One example is a data visualisation by Kamel Makhloufi where he showed the civilian deaths in the Iraq war as an area chart. The interesting aspect is that each pixel represents exactly one person. So you have a one-to-one relationship to the data, rather than an abstraction. It looks just like any other area chart, but there is a direct relationship between what you see on the screen and the raw data.

When working with data in this way, your intention is to help the viewer, or recipient, piece the story together based on traces. I may have started with data visualisation, but I'm much more fascinated by the idea of mapping, rather than things like network diagrams, because there's so much richness and culture in it. People like Rob Kitchin talk about the map as something that the recipient has to go through. The creator cannot just say, “Okay, this is the representation. Try to understand it.” Instead, they guide the recipient through the process of piecing something together.

Angela —
So working with data is not always a linear, or prescribed process?

Dietmar —
I think working with data is like being a designer. If you take data as a material artefact seriously, there are so many different material practices that you can apply to it.

Reservoirs of Venice by Dietmar Offenhuber and Orkan Telhan treats the city as a computational system, using its physical dynamics to predict environmental change. Screen Column draws live data from Venice’s canal webcams. Photo: Sebastian Gonzalez Quintero, May, 2025. Videos courtesy of @iloveyouvenice.

Autographic design is a process of framing traces in data. There are always more traces than you can consider. Many are invisible; some are missing. The most conspicuous traces are not necessarily the ones that are the most meaningful. There are groups who engage in these practices of autographic design. They frame these objects and trace phenomena in a certain way to explain what is going on. One other example is what Giorgia Lupi and Stefanie Posavec did with the Dear Data Project. It's not about making cute pictures of data; it's about sincerely recording and generating data in a systematic way that is constructed by the participants.

Autographic design as a practice involves framing, but also recording, registering, aggregating, and encoding. The processes and stages of data generation have a certain experiential quality, and through autographic design we become aware of the common pitfalls of data analysis that the literature of critical data studies cautions us about. In autographic design practice, the traces of data generation are revealed and shared as a public process of evidence construction.

Another example is when we work with data that is recorded by 9-1-1 police call-centres. When people phone in an emergency, the responder types the conversation into a keyboard, and of course, they make typos or spelling errors. These mistakes are kind of autographic, they are elements that sometimes allow you to identify the different people who have written those reports.

GISP2 ice core section showing annual layer structure (cropped), illuminated from below. Source: ftp://ftp.ncdc.noaa.gov/pub/data/ paleo/slidesets/icecore-polar/

Angela —
Can an autographic design lie?

Dietmar —
There is this almost moralistic idea that you should not lie with a material. This was very strong in architecture, and it is also a topic in scientific representation. Lauren Daston and Peter Galison have worked extensively on the ethics of scientific images. For architects, it is much more practical. Whenever an architect draws a line or outline of a building on a plan, it's never just a line. It always has a body. It always has a width, and that body always generates problems. And so, architects must always wrestle with the materiality and the digital space.

Angela —
What about the value of openness?

Dietmar —
It's important to have data open. That's not a question. A society needs the ability to scrutinise its own data. My own country, Austria, still doesn't have a freedom of information law. You still have arcane laws from the monarchy that enable the administration to maintain absolute secrecy about their processes. You would not think that this still exists, but it does. I think the US Freedom of Information Act had a very positive impact in many ways.

But we have many examples of how misinformation takes advantage of open data. It's very easy to use open data as a basis for denying climate change, to misconstrue certain events, or attack enemies. There's also the aspect of refusal. Since you're a Canadian, Angela, I will use the example of certain First Nations who refuse to be counted in census or some data collection processes. That right to refuse is not compatible with the traditional thinking about open data. We must understand that there are situations where the availability of data can be harmful or can be distorting, and it can be an issue of power. James C. Scott talks about the state that has standardised measures that are not nuanced, but that allow integrating information over a very large area, whereas the people on the ground have different modes of measuring that are sometimes conflicting. People don't always want to be revealed in those kinds of state representations.

Angela —
And of course, there is also the issue of personal privacy around data that can reveal one’s identity.

Dietmar —
I see that there are a lot of ethical frictions in this topic due to the trade-offs between privacy and data utility. The US census has recently decided to only distribute synthetic data sets. The population counts have noise injected, and the microdata include processing so you can't link them to individuals. Social scientists are, of course, unhappy because they say they can't really do exploratory data research anymore. If the data is machine-manipulated, you are no longer able to engage with the dataset from a purely analytical perspective, but you have to treat it as some kind of embodiment that comes out of a deep learning model.

Angela —
And then again, you look for the traces of the machine in the data?

Dietmar —
Even the AI, because it has bias, it's not a personal bias, it's the bias of the machine or the model. In any case, we're forced to adopt the relational concept of data where we must consider the situation, the application, and the circumstances, rather than summarising everything with a representational relationship.

Angela —
Should we then improve data literacy?

Dietmar —
In the US, especially starting in high school, data literacy is a very big topic. The idea is to make sure that students can understand data and data visualisations. Of course this is important, but by doing this, you are implying that there is a standard way to create and then decode a data visualisation. But this is not how people make sense of visual phenomena. We construct meaning based on what we see. People can find a certain meaning in an image even if they do not understand anything about it. It might be completely wrong, but everyone will have an impression. This is not the same as being unable to read, where it is impossible to understand a sentence if you cannot decode the letters and words.

In terms of post secondary education, data visualisation is often in computer science departments, but they often miss the design part of the process. We need the technical aspects but also how to engage with a problem, how to iterate, and how to test a data design. It should connect the computational dimension, the design dimension, and the humanistic scholarly dimension.