Representative Data

What, or who, gets researched is incredibly important to consider. We cannot make life better for people if we do not understand the nuanced experiences of different people. We cannot understand the world if we only focus on one aspect of it. One way research is biased is through the selection of data we collect and analyse. This is relevant for primary (data we collect ourselves) and secondary (data collected by others that we analyse) studies. A fundamental part of any research study involves a certain amount of data exclusion; deciding not to interview participants with certain characteristics or removing data points outside of a certain range. Whilst some data curation is usually necessary, it is important to think carefully about the consequences of these choices on how representative our results are. As argued in the article ‘Why Data is Never Raw’ by Nick Barrowman, all data reflects, to some extent, the pre-existing theories, assumptions, interests, and limitations of the researcher who collected it.

Data can directly represent human experience or aspects of bodily functions (such as diary excerpts or blood pressure measurements), but it can also indirectly reflect our motivations and interests (from spending habits to observations of the universe). Wherever data is collected from, it is important to consider who it represents and who it benefits, whether directly or indirectly.

Historically, research has overwhelmingly represented the identities and interests of a global minority: white cis-heteronormative men in the global north. Despite the research environment becoming more diverse in recent years, current practice and institutional structures largely continue this legacy. Whether it is the histories that get told, the theories that get taught, or the data that is collected, research and our understanding of the world have been disproportionately shaped by this global minority identity. Certain human characteristics, geopolitical regions, species, and ways of knowing are under-recognised or excluded in research.

One of the issues with data representation is that the characteristics of participants or samples are often not appropriately recorded. The reasons for this can be manifold, though often it is simply the result of oversight (potentially due to unconscious bias) or belief that certain characteristics are not relevant to the research. As argued in ‘Four ways natural history museums skew reality’ by Jack Ashby, absence of data does not mean it does not exist, or that it is not important. Furthermore, this Guardian interview with Catherine D’Ignazio: ‘Data is never a raw, truthful input – and it is never neutral’ discusses the harm these gaps can produce when data is used by algorithms and AI.

Our page on Co-creation and Collaboration introduces some ideas for how to engage with communities in a way that improves representation. However, caution must be taken not to overuse certain participant groups, called ‘convenience sampling’. Not only does this fall back into the trap of over-representing a particular population, but it can also induce involvement fatigue within this group. Given that such groups may already be marginalised and experiencing exploitation within society, researchers must take care not to compound such problems. Instead, it is important that we challenge ourselves to find new representative groups and therefore broaden our understanding rather than narrow it.

Representative data is crucial for research that seeks to accurately represent the world and yield positive outcomes. Global power dynamics act to exclude certain groups of people, places, concepts, and phenomena from data collection which limits the validity and usefulness of research. Researchers must critically assess their data choices, seek broader representation, and engage with communities ethically—without relying on convenience sampling or reinforcing existing inequalities.