FUNDR DAILY NEWS
SOCIAL NEWS FOR ALL TO READ
Why charities need to go back to basics to benefit from AI and data science
This article was written by Giselle Cory and Tracey Gyateng, DataKind UK – the organisation that offers charities free support to transform their impact using data science.
Giselle Cory and Tracey Gyateng, DataKind UK, argue that charities should look beyond the big scary concepts of AI and data science and instead look at the simple things they can start putting in place now that will answer their most important questions.
AI is going to be transformative for human civilisation. Everyone should benefit from it— and be protected from any harm it may bring.
That’s all well and good, but it’s not the most pressing data problem that social sector organisations bring to us. In fact, it rarely even makes the shortlist. The glamorous digital dreams of transformative AI are still far away. This is true of the private sector and the social sector, which tends to be slower to adopt technological change.
But aren’t the robots coming?
Not just yet! What is known as ‘artificial general intelligence’ – a machine capable of true learning, rather than heavily context-specific learning – has not been developed yet. Most AI advances are in the area of ‘narrow AI’ – where machines are very good at a single task, like playing chess.
Data science uses a mixture of computer science and statistics to draw insights from data, often deploying a range of narrow AI techniques such as machine visioning (identifying images) or machine learning (finding patterns in the data).
However data science is a lot more than just AI. (If you want to know what is and isn’t AI, this flowchart might come in handy). In fact, AI isn’t even the most useful bit of data science for many social sector organisations. For many, it’s the more vanilla bits of data science that are also the more insightful:
- Descriptive analysis — who is using the service & how?
- Assessing relationships (or regression) —what other factors (like someone’s age) are associated with good outcomes from using the service?
- Profiling — which groups of beneficiaries respond best/worst to the programme?
- Data quality — Is my data any good in the first place?
- Combining datasets — Can we join different datasets together to see if the beneficiaries from one programme overlap with those from another? (This is part of a process that is broadly called ETL — extract, transform, load)
That’s not to say that the social sector gets the boring bits while the private sector does the flashy AI. The private sector heavily relies on the techniques above, too. And even when firms say they are using AI, many aren’t. Moreover, it’s often impossible to do decent AI without doing these other bits first – particularly getting your dataset in order.
Basic approach, big impact
These techniques can be immensely powerful for social sector organisations. For corporates, they can help to micro-target you with a nice pair of sandals. When used responsibly within the social sector – which means taking into account ethical as well as legal considerations – data science can help vulnerable people to get the support they need.
Sometimes it’s the straightforward analyses that are the most transformational for an organisation. Not sure if your programmes are working for young people? Want to know if 6-month programmes are more impactful than 3-month ones? You need data science, not AI.
More than data
All this talk about data science can overlook the biggest challenges for social change organisations trying to become (more) data-driven. For many, the actual data science is a small piece of the puzzle. What about getting senior leaders to pay attention to the results? And how do we even store this data in a safe and responsible manner?
It may sound counterintuitive, but when we partner with social change organisations , the main aim isn’t to find definitive answers to their questions. Yes, it’s great to see the accessible, insightful results that the teams produce. It’s even better to see the transformation in the organisation’s understanding of data science. It’s this increase in their data maturity that we want to see. It means that they are learning and changing, rather than getting helicoptered-in answers they can’t interrogate, reproduce or embed.
Bring the boss
A key element to support this long-term change in data maturity is getting senior leadership involved. Without that senior level buy-in, there’s unlikely to be any investments in data science capacity. That’s why we celebrated the participation of senior leaders in our July DataDive. Our charity partnerStreet League brought their Director of Strategy and Impact, and The Mixteam were joined by their CEO.
So – talk about AI and what it might mean for your organisation. But put the foundations in place first. See what your data can tell you from some straightforward interrogation, consider if your data plumbing needs fixing (how you store your data) and see if you’re collecting the right data in the first place.