Fully Funded Swansea and No More Marking Ltd PhD Scholarship: Elevating Comparative Judgment Using a Large-Scale Human-In-The-Loop Bayesian Active Learning Approach (ECSTATIC)
The Comparative Judgment (CJ) method, popular in UK schools, involves selecting superior submissions from pairs. A Bayesian active learning approach (BCJ) has been developed to improve interaction-efficient pair selection and predictive uncertainty. The project aims to scale BCJ to handle thousands of items, propose new methods for dynamic ranking, and provide individual learners with progress tracking. The methods will be designed in collaboration with assessors and learners.
Scholarship Summary:
Qualification Type: PhD
Location: Swansea (UK)
Funding available for: UK Students, EU Students, International Students
Funding amount: £19,237 p.a.
Hours: Full-Time Research
Last Date of Application: 1st May 2024
Funding Source: Swansea University’s Faculty of Science and Engineering and No More Marking Ltd.
The subject of Study: Human-Centred Artificial Intelligence, Optimisation, and Decision-Making
Project Description
Over the past ten years, the Comparative Judgement (CJ) technique has gained popularity in UK schools. Instead of giving each submission a mark, assessors select the better one from a pair. For a limited number of entries, this method preserves accuracy while putting less strain on assessors. We recently devised a Bayesian active learning strategy for CJ (BCJ; https://arxiv.org/abs/2308.13292), which yields trustworthy estimations of rankings and predictive uncertainty while solving the important challenge of interaction-efficient pair selection.
In this linked project, our goal is to scale BCJ to handle thousands of things (instead of tens of them) for the first time, so that scoring and ranking can be done across assignments and schools. We will suggest fresh approaches to dynamically add new ranking items in BCJ in a way that maximizes engagement, and we will come up with ways to provide individual students a sense of how they are doing over time in relation to their classmates. We will analyze these approaches to determine how well they teach learners and assist assessors in making decisions in the face of uncertainty resulting from the practical lack of data and interactions. These techniques will be created in cooperation with assessors and students to make sure they are applicable and helpful once the project is over.
Eligibility
- You must hold an Upper Second Class (2.1) honors degree or master’s degree with a minimum overall grade at ‘Merit’ in Computer Science, Mathematics.
- English Language: IELTS 6.5 Overall (with no individual component below 6.0) or Swansea University recognized equivalent.
Desirable skills and attributes:
- Excellent numerical and programming skills;
- Knowledge of Python.;
- Knowledge of Bayesian statistics, machine learning, and optimization, or a willingness to learn.
This scholarship is open to candidates of any nationality.
Additional Information
This scholarship covers the full cost of tuition fees and an annual stipend at £19,237.
Additional research expenses will also be available.
