Following Anna's blog I have also watched the Cambridge Analytica documentary, The Great Hack. It was sobering to consider the physical and verbal violence perpetrated during elections around the world due to the manipulation of people's personal data.
It seems to me to be vitally important that the social impact sector grasp the significance of these technologies in order to utilise them for good. This excellent SSIR article https://bit.ly/2yPR08G points to applications of machine learning within International Development. Machine learning is a subset of AI and uses statistical techniques to teach computers to carry out actions without human programming. The author highlights four different types of machine learning. Below I summarise them and their potential applicability across all non-profit sectors, point to some potential examples and highlight questions to see if you can find resonance with your organisation’s mission.
1. Natural language processing
Do you work in a sector or organisation where there are large text based data-sets available but left unanalysed?
Natural language processing can efficiently review and consolidate text data, draw inferences and identify trends that humans just haven't had the capacity to extract.
Donors, for example, may have a wealth of information submitted to them about a sector, a problem or a location, that never gets utilised or analysed because applicants were unsuccessful. This could provide critical contextual information to guide future programmes.
We often try and analyse the needs of the communities that we serve in greater depth. Cluster analysis goes beyond segmenting a target group by sex, age or location. Given the right data set, it can also provide insights about behaviours, language, habits, preferences and decision-making criteria.
For example, this could be utilised to encourage a specific community to become more eco-conscious in their purchase decisions by using a language and approaches that are familiar and more likely to facilitate behaviour change. Or cluster analysis could help ensure that an intervention is made at the right time and in the right way to support an at-risk individual who is struggling with mental health issues.
What would you like to understand in more depth about your communities? What difference might that make to your impact?
3. Predictive machine learning
Predictive machine learning helps us anticipate what will happen. ‘Classification’ analyses available data and thereby allocates somebody or something to a certain group - for example, a non-profit could identify which kids with literacy issues are most at risk of becoming involved with gangs due to other life factors and ‘Regression’ predicts a specific quantifiable indicator, such as the grade a particular student is likely to attend in their exams.
Predicting outcomes or behaviour is something that we humans are notoriously poor at, the opportunity to intervene or mitigate these outcomes is potentially life-changing.
How would it help your mission if you could predict the future?!
4.Causal machine learning
Causal machine learning lets us identify linkages and underlying causes between different factors, and once we define this explanatory structure, it enables predictions about behaviour if any of those factors are changed. For example, what would happen to the number of girls forced into early marriages if awareness-raising lessons were introduced at the age of 7? Or 11? Or 13? This is systems thinking on steroids and could allow us to manage a huge level of complexity with greater ease and effectiveness.
At the moment most non-profit organisations have a threefold issue; we don't have the staff with appropriate knowledge; don't have the resources to fund the data analysis; don't have data-sets that are comprehensive enough to utilise the power of machine learning. Oh and maybe a fourth issue, we don't have the time or resources to even explore these options because we are too busy with The Day Job (Been there, got the tee-shirt...)
But what if we put just a few hours aside to start thinking about these possibilities and taking baby steps so we’re fast off the line when the funding comes? It will.