An exhibition exploring how data is shaped, curated, and made meaningful
This exhibition reflects on how data becomes knowledge through processes of selection, organisation, and presentation. What often appears as a clear answer or neutral decision is actually the result of layered decisions all about scale, categorisation, visual form, and inclusion. These shape how information is understood.
Moving from broad representations to the fine more niche elements, the work traces how meaning is produced at different levels of abstraction. Each step reveals how data is framed: which indicators are foregrounded, which comparisons are enabled, and which uncertainties are smoothed out or concealed.
With our understanding of data from our discipline of cognitive science, we wanted to ensure that our exhibition fit our target audience, other cognitive scientists. We questioned the approach to the assignments and decided against chronology and focused more on the aspects of the course and data in regards to what they entail.
By slowing down the encounter with data, the exhibition invites viewers to question the authority of seemingly seamless outputs and to consider how knowledge is constructed, mediated, and made to appear self-evident.
You arrive at the Big Picture – A polished view a user might see when they ask the system a question. At this layer, the messy work of collecting, structuring and storing data has already happened in the background, with the user having no knowledge of the process. What remains is a seemingly neutral answer, right?
Imagine a user request: a search term, a filter, a dashboard, or a simple question. The system responds with a specific view of the world; a map, a graph, a ranking, a story. This view feels like a presentation of real world data, but it is already a constructed perspective: someone decided which variables matter, which are combined, and which are left out.
At this level the underlying processes are hidden. The interfaces do not show where the data comes from, how it was cleaned, which categories were used, or how uncertain the numbers might be. The process from raw traces to this final output is effectively a black box, inviting us to trust the result without seeing the steps that produced it.
Even here, curation is at work. The Big Picture hides as much as it reveals: it selects which indicators are visible, which comparisons are possible, and which questions can be asked at all. Design choices such as colours, scale, labels, default views, quietly guide interpretation. Bias is not only in the data itself, but also in how the answer is framed for the viewer.
Stepping down from the Big Picture, you enter the visual and interface layer. Here, data is translated into charts, maps, diagrams and other visual forms that make it readable and sensible.
Visualisation converts data into shapes, lines, maps or patterns that the eye can interpret quickly. Good visualisation highlights structure without distorting meaning, helping the viewer see relationships that are difficult to grasp in raw numbers.
Visualisations come from specific positions: decisions about what to collect, how to classify, and what to display reflect the worldview of the designer. Critical approaches remind us that visualisations are never neutral — they foreground some perspectives while silencing others.
Visualisation also tells stories. Choices of colour, scale, comparison, and layout guide how viewers interpret meaning. Some visualisations aim not only to inform but to create a felt, experiential response — making data more embodied, emotional, or atmospheric.
These files document the visualising & interface work behind this level.
Going one layer deeper, you reach the infrastructures that make those interfaces possible: storage systems, databases, and archives that hold and deliver the data behind the visuals.
Data lives in physical environments, which are servers, cooling systems, fibre routes, warehouse-scale data centres. Storage is shaped by geography, energy, and industrial infrastructures, which decide where data can exist and at what scale.
Cloud infrastructures distribute data across borders, blending technical convenience with political complexity. Control over storage becomes a question of sovereignty, corporate power, and who gets to set the rules for access, ownership, and protection.
Archives decide what is preserved, forgotten, or made searchable. They shape cultural memory by choosing how materials are described, contextualised, or restricted. Storage here becomes ethical: whose histories are kept, and under whose authority?
Infrastructure is never neutral: design choices, metadata practices, server lifecycles, and retention policies shape what data even survives to be categorised or visualised later. Bias begins here; before datasets, before interfaces but in the foundations of storage itself.
Below infrastructures lies the semantic layer, where categories, vocabularies and Linked Open Data structures make the dataset machine-readable and connect it to wider knowledge graphs.
Categorisation is where meaning starts taking shape. The way things are grouped, separated, or named influences what patterns become visible and what stays hidden. Classifications are never neutral, they actively structure the world they claim to describe.
Linked Open Data connects individual entries into larger networks. Projects like Wikidata show that data modelling is both technical and social: properties, links, and entities express shared assumptions about what counts as knowledge, and whose knowledge becomes part of the official record.
Choices made during categorisation embed bias long before data becomes a dataset or a visual. Selecting properties, defining classes, and deciding which perspectives belong in the structure all shape what is thinkable and searchable. At this level, curation is about deciding what relationships are possible.
These files document the categorisation and Linked Open Data work behind this level.
Deeper still, you reach the dataset layer itself: collected objects, records, and possibly synthetic data generated to extend or protect the dataset.
At this layer, the world is translated into records, fields, and counts. Everyday objects and events become data points through processes of selection, measurement, and encoding. What looks like a simple dataset is the result of many small decisions about what to capture and how to standardise it.
Synthetic data is generated rather than directly observed, often to protect privacy, augment scarce datasets, or simulate scenarios. It promises more data without more surveillance, but also opens new markets where data itself becomes a product, raising questions about power, value, and control in data-intensive capitalism.
Bias enters through what is collected, what is left out, and how records are cleaned, merged, or even synthetically generated. Metadata, formats, and quantification practices frame what appears normal or exceptional in the dataset. Here, curation is about constructing the very raw material that later levels treat as “given”.
These files document the collecting and dataset construction work behind this level.
At the deepest level, you arrive at data itself: individual data points, bits, signals, and the physical infrastructures and uncertainties that surround them.
At the core, data reduces the world into discrete points: numbers, signals, coordinates, attributes. These points only become meaningful through context, interpretation, and prior decisions about what should be captured. Nothing is inherently raw, even the smallest values are curated abstractions of reality.
Even a single data point reflects choices: what was measured, how precisely, under what conditions, and with which tools. Bias enters through sensors, thresholds, missing values, and assumptions hidden in the smallest units. The core is not neutral, it carries traces of the systems that produced it.