How Can AI Help Underrepresented Young People Get Into the U.K.’s Top Universities?

Use Cases & Projects, Dataiku Product, Scaling AI Valentine Reltien

Dataiku’s AI for Good program, Ikig.AI, provides free Dataiku licenses to accelerate nonprofits’ missions. These collaborations often reveal outside-the-box use cases with tangibly positive human impact. This blog post will walk you through Ikig.AI’s AI partnership with The Brilliant Club, a U.K.-based nonprofit seeking to increase social representation in prestigious universities. 

1. A Social Justice Issue That Bridges Nonprofits & Corporations

Dataiku cares deeply about advancing matters of Diversity, Equity, & Inclusion (DEI) in the tech space — hence why it is one of Ikig.AI’s three avenues of actions. Dataiku is well aware that it is a matter that requires systemic change. Better representation of diverse backgrounds in corporate tech relates to representation of students pursuing the subject matter in higher education. Research has proven that this is connected to the quality of students’ earlier education which itself results from pupils’ socioeconomic status and the school they attend as a result of the former.

When The Brilliant Club and Ikig.AI were first introduced in early 2021, we were compelled by the opportunity to get to the root of the problem of misrepresentation in the tech space. Better still, thanks to The Brilliant Club’s Research and Impact Department, the organization was particularly mature and well-equipped with responsibly collected data to reap swift benefits from an Ikig.AI AI partnership. 

In England, only one in 33 young people from disadvantaged socioeconomic backgrounds progress to university education, compared to one in three in most advantaged backgrounds. The Brilliant Club exists to increase this number and does so via its Scholars Programme. Therein, doctoral researchers are placed as tutors in schools to deliver university-style tutorials and two trips to university — one before the program and one upon completion of their final assignment. Together, these three elements effectively pave participants’ way to university, thus sealing the success of The Brilliant Club’s mission.

young adults in classroom

2. Beyond Shared Values and Suitable Tools: How Could Dataiku Enrich the Scholars Programme?

Let’s back up. What was the matter to begin with?

In spite of The Brilliant Club’s polished and proven Scholars Programme, in 2020 12% of over 8,000 participating pupils did not submit their final assignment, meaning they did not complete the program. These final assignments are academically challenging: They are often the longest pupils have written at that point in their education. Those who complete them are invited to graduate and celebrate their achievements at one of The Brilliant Club's partner universities. Considering solid evidence that proves the program’s positive impact on graduates, The Brilliant Club knew it was of the utmost importance to explore factors underlying the cause for not completing final assignments.

The Brilliant Club was swiftly onboarded as a pioneer of Dataiku Online. Given its light infrastructure requirements and rapid registration, The Brilliant Club was quickly able to access Dataiku’s platform value-generating features. Their onboarding was made even more conducive thanks to Dataiku’s plentiful onboarding trainings and the assistance of a volunteer Dataiker data scientist using Ikig.AI days for pro-bono work.  

The team at The Brilliant Club already responsibly collected data from over 70,000 Scholars Programme pupil records (which included demographics, engagement data, school and tutor-level information), so they were quickly able to process and clean data in order to move onto modeling. They then hoped to harness Dataiku’s platform predictive modeling capabilities to better understand what was holding their mission back.

How did we tackle it?

Ultimately, The Brilliant Club sought to output a “risk rating” — namely, a prediction of whether a pupil had a high, medium, or low risk of not submitting their final assignment. This part of our collaboration with The Brilliant Club highlighted firmly held values at Dataiku: A model can never run on probabilistic relations alone. To best devise a successful model, The Brilliant Club’s Research and Impact Department — a team of social scientists familiar with the on-the-ground reality of the program — was consulted to carefully select key qualitative data features to drive final assignment submission and non-submission.

Thanks to Dataiku’s refined explainability features, The Brilliant Club was able to gather and easily communicate granular insights from the gradient-boosted decision trees (which resulted in optimal accuracy — 81% — between test and train datasets) with non-coder colleagues. This aspect of our collaboration was crucial to The Brilliant Club as this modeling exercise only made sense with the perspective of implementing actionable steps from qualitative data. What is more, the transparency of the visual flow and ease of understanding were essential to ensure the model’s maintenance and replicability so that value should continue to “flow” from this one project on to refreshed datasets and different users.

3. Turning Ideal Aspirations Into Actionable Insights: Dataiku's Model #IRL

Following this collaboration, The Brilliant Club’s Research and Impact Department returned to its program management teams to integrate risk ratings in their everyday operations. Putting this model’s predictions to use required on-the-ground innovations to better implement preventative measures for those pupils at high and medium risk. In addition to raising their teams, schools, and tutors’ awareness of risk ratings’ constitutive factors and traceability, The Brilliant Club set two points in each termly program to calculate these risk ratings and thus reassess students’ conditions.

risk ratings The Brilliant Club

Risk ratings and ensuing interventions to support medium- and high-risk students were fully implemented in The Scholars Programme as of the spring term of 2022. In great data-driven fashion, The Brilliant Club’s impact team are intent on experimenting with different intervention strategies (i.e., offering additional tutoring in smaller groups, raising teacher awareness for more support, dedicated tutor trainings, etc.) whilst tracking their effect on pupils’ ratings from one term to the next.

The demonstration of this data analytics and predictive model’s positive impact on The Brilliant Club’s missions will be reflected in the improvement of two key metrics decided on by The Brilliant Club’s teams:

  1. The percentage of pupils identified as ‘high risk’ decreases between rating 1 and 2 as a result of interventions put in place after rating 1.
  2. The percentage of pupils identified as ‘high risk’ who do go on to submit final assignments has increased with the introduction of risk ratings (e.g., comparing 2020-21 with 2021-22).

Ultimately, continued data collection on attendance and assignment completion is still crucial to assess optimal risk mitigation strategies. Fortunately, Ikig.AI and The Brilliant Club’s partnership is one on the long term which enables us to follow up on this project’s evolution and the effect of implemented measures.

The Brilliant Club’s ability to harness AI to better deliver their mission thanks to Dataiku's accessible, transparent, and responsible platform is a collaboration Ikig.AI is humbled and excited by. Beyond fulfilling Ikig.AI’s ambition to advance DEI in our local communities, this partnership highlights the creative potential and positive social impact AI can assist in. Check out this blog post for further information on how Ikig.AI by Dataiku supports nonprofits’ missions. 

You May Also Like

Alteryx to Dataiku: Working With Datasets

Read More

Demystifying Multimodal LLMs

Read More

I Have AWS, Why Do I Need Dataiku?

Read More

Talking AI Democratization With Dr. Anastassia Lauterbach

Read More