Insights
The 2021 Data4Good Festival demonstrated the need to look at data through a different lens. We explore how data is not separate from power structures and focus on its relationship to equality, diversion, and inclusion
The Data4Good Festival 2021 was hosted from 10 – 12 May 2021. The festival shared inspiring stories of how social sector organisations have used data to improve impact. The talks varied, from examples of data journeys charities have taken to workshop ‘how to’ sessions.
Attendees were able to find talks through filtered themes and it was refreshing to see one titled ‘Data and Diversity, Equality and Inclusion’. This theme stood out, because it positions data as something fixed within our society and culture, not outside of it.
Several sessions helped us reflect on the ways we collect data – such as the process of collecting Equality, Diversity, and Inclusion (EDI) information from users and staff. Sian Basker from Data Orchard encouraged organisations to compare the protected characteristics to the data they currently have on staff and users. You’ll probably find that you have more complete and up-to-date data on one characteristic (such as age) but not enough on another (such as marriage and civil partnership).
Even within these protected characteristics, there are questions surfacing around their definitions. As our culture changes and we begin to scrutinise the language behind these terms, collecting data about DEI is a process that needs reviewing. Kevin Guyan from EDI Scotland hosted the session ‘Count me in – the collection of diversity monitoring data and its use for action’. This session exposed how diversity monitoring questions should be formatted in a way that enables users and staff to identify accurately.
For example, the only protected characteristics that allude to gender identity are ‘sex’ and ‘gender reassignment’. However, this still excludes a group of people who do not identify with the sex assigned to them at birth, but who also have not gone through the process of gender reassignment.
You can ensure data collection on the protected characteristics is inclusive by providing more options to answers. By having a better understanding of what characteristics such as ‘disability’ encompass, you’ll be able to collect more accurate and representative data of users and staff.
Another session that was thought-provoking was the panel discussion ‘Decolonising data: principles for improving the ownership, diversity and accessibility of data’. The session drew on similar points to others about how data collection reflects biases and social inequality. Certain research methods – particularly when collecting qualitative data – may only reproduce power structures both within the community and globally.
For example, when collecting the narratives of locals in a community that do not speak English, researchers rely heavily on the interpretation of translators. Furthermore, for the members of the community who can speak English, they may be involved more in the research process than other locals.
This reinforces dynamics within a community where English-speaking members may have more power, status, and wealth than others. The panel also felt that analysis and findings should be worked on and shared with research participants. Nonprofits should share their research with the communities they are working for, not just with funders and supporters. Findings should be actionable and produce change for these communities.
The final keynote speaker, Rachel Coldicutt from Careful Industries, described data as “sociotechnical material”. Data reflects the interests of people collecting it and inherits power structures existing in our society. Rachel’s examples showed how data can perpetuate forms of social inequality such as racism.
For example, the Metropolitan Police ran an algorithm that estimated the likelihood of young black men at risk of being in gangs. What the algorithm couldn’t differentiate between was someone who was a victim of crime or a perpetrator of crime. The characteristics of young black men were muddled with characteristics of someone in a gang, and the consequences of being on the matrix could affect access to housing, work, and benefits.
Rachel also spoke about a scandal that occurred in 2018 about an insurance company using names as an indicator of risk for car insurance. The Sun reported that people named Mohammed could pay £1,000 more to insure their cars than someone called John, for example. While the insurance company denied the accusation, this demonstrated ways we frame data can mirror biases towards groups of people.
The EDI sessions proved that data is not objective, but can be collected, analysed, and reported to serve a person’s or organisation’s agenda. However, we can produce more representative and inclusive data when we question our methods and inherent biases.
The Data4Good festival was a great opportunity for the charity sector to question its approach to data and its impact – both negative and positive.
Join us on the 14th of May for our Q&A session. It will provide a whistlestop tour of Microsoft Copilot’s key capabilities, how they can help charities, and answer all your burning questions around Microsoft’s AI service.