Martin Lewis Fund for Charities Used Algorithm To Skim Applications
The emergency fund set up by Martin Lewis to support charities during lockdown used a machine learning algorithm to skim the high number of applications it received.
On 19 March, Martin Lewis, the founder of comparison website Money Saving Expert, launched a £1m fund to help small charities during the crisis.
The fund closed on 25 March after receiving 7,000 applications in six days, totalling about £74m.
An algorithm looked at keywords in applications
Jonathan Cook, a fundraising consultant who worked as charity lead for the fund, spoke at the Chartered Institute of Fundraising’s Convention about how the team reviewed the applications in order to distribute the money to charities as soon as possible.
He said: “Martin was absolutely adamant that one week after opening, he wanted the first batch of grants to go out. So we made it quite an automated process involving a very simple application form done online with about five boxes.
“In order to plough through this enormous number of applications, we used some computer algorithms, which picked out key phrases and keywords that people can use in their applications. Phrases like ‘free school meals’ – if we discovered that phrase in an application, it drew attention to the fact that it might be talking about provision for people who are now no longer at school and might have a problem feeding their children.”
“We also looked at the locations [where the applications came from] – whether they were in areas with high incidence of Covid-19, whether they were in areas of high deprivation. That enabled us to focus our attention on a smaller number of grants and we could then have a human eye look through them.”
Cook also said the use of a computer algorithm removed potential issues with people’s unconscious bias, at least at that initial stage.
However, Fozia Irfan, CEO of Beds & Luton Community Foundation, who was speaking at the same panel discussion, pointed out that algorithms present their own issues in terms of equality, because they can reflect the unconscious bias of the people who originally wrote them. This is a phenomenon known as “algorithmic bias”.
The fund, which in the end amounted to £3.4m thanks to the support of other donors, had been distributed in its entirety by 26 May to a total of 415 organisations.
It focused on projects providing immediate relief during the crisis, prioritising “the provision and delivery of food, medicine, sanitary products and emergency hardship grants”.
Read the full Civil Society article by clicking here.
Source: Civil Society