3 days ago
AI to help mark student exams
Artificial Intelligence is to help grade A-level and GCSE exam papers.
Exam board AQA is working with King's College London academics to create an AI virtual assistant to assist and support exam paper markers.
More than £600,000 of taxpayer money is funding the three-year scheme which intends to boost the accuracy and reliability of the grading process.
Professor Yulan He, an expert in using machine learning to understand written text, is leading the work and said the aim is to reduce errors, make mark schemes fairer, and give quicker feedback to students.
'We aim to create an AI technology that can add a layer of quality control and support experts mark accurately and consistently,' Professor Yulan told The Telegraph.
'The goal is very much not to create an AI system that replaces examiners marking students' work, but recognises that marking is a really demanding task and supports them to mark more consistently and faster.'
The AI assistant could be used to check the marks given by a human and detect any scores which seem erroneously low or high based on the AI's own assessment.
Other possible applications are to check the quality of answers from students using language analysis machine learning and scrutinising the relevance, factuality, coherence and logical reasoning of an answer.
The AI could also be used to refine the mark scheme if there is a flaw in how marks are given to ensure fairer marking and also to give AI-generated explanations to students as to why they did or did not get a question correct.
Government backing for AI in education
Peter Kyle, the Science and Technology Secretary, told The Telegraph the project will help pupils reach their potential and 'open as many doors as possible'.
'The central, most exciting, opportunity that AI affords us is not where it can replace human expertise, but where it will enhance it,' he said.
'By adding a layer of support and quality control for exam markers, this work by King's College London can make sure every student receives the fair assessment they deserve, to fulfil their potential and open as many doors as possible for their future success.
'These projects are combining cutting-edge research with professional expertise to turn research into practical solutions that can improve people's daily lives and drive growth as part of our Plan for Change.'
Prof Yulan insisted the human involvement in marking papers is critical to the process and AI will be used to facilitate and refine instead of to replace.
'This is not to replace the human marker. Humans in the loop are very important. Our AI technology will just add a layer of quality control and support experts mark accurately and consistently,' she said.
'There is no danger of this being used to mark next year's exam papers unsupervised, for example.
'We aim to develop a safe and reliable virtual assistant in collaboration with teachers, learners and subject experts, that could be used in the future to assist examiners when they mark.'
Pepe Di'Iasio, general secretary of the Association of School and College Leaders, said: 'If AI can provide an extra level of quality assurance, which supports examiners and improves the accuracy and consistency of grading, then that is clearly a good thing which we very much support.
'As with any project to utilise AI, it is absolutely essential that it is robustly tested before being rolled out into the real world, and that exam scripts are always seen by human beings. We're pleased to see assurances to this effect.
'The process of marking and grading exams is a major undertaking which has to be delivered to tight deadlines, so the potential of technology to support this process is certainly something which should be explored. We look forward to seeing how this project develops.'
'People should always remain in control'
AQA is one of the country's largest examination boards.
Around 1.4 million students sat one of its qualifications last year and more than nine in ten schools have at least one AQA course in its curriculum.
Around 30,000 teachers, lecturers, subject experts and academics help design its tests and mark exams each year.
Cesare Aloisi, AQA's head of assessment innovations, said: 'We're always looking for ways to enhance the support we give the examiners who mark our exam papers and to develop even more sophisticated quality controls.
'AI can help us do that with the important guiding principle that it is there to help human expert judgment, and that people should always remain in control.
'For example AI could flag marks that may be too lenient or too severe, but it would always be a person deciding what mark to give. We think it is essential that our expert examiners, and not AI, make these decisions that are so important for young people's futures.'
