Intelligent Advisory System for University Admissions

Software	Any good language for developing AI systems, or possibly an expert systems shell
Covers	Artificial intelligence, probabilistic thinking
Skills Required	Programming, interest in artificial intelligence, ideally some interest in statistics.
Challenge	Conceptual ??? Technical ?? Programming ????

Brief Description

The objective of this project is to develop an intelligent system that takes students' exam results and other information, as supplied on their UCAS forms (as they are now, or with whatever enhancements you think UCAS forms should have), plus any tests the university requires or administers itself, and makes recommendations for university admissions decisions. This could include setting the levels of conditional offers.

Faced with applicants taking a plethora of different entrance qualifications, university admissions tutors and administrators (notably at De Montfort University) find it extremely difficult to predict who is likely to do well at university. It is a constant source of concern to universities that they might (without realising it) be making offers demanding different standards of attainment to applicants doing different qualifications, such as A Levels, Scottish Highers, Advanced GNVQs, BTEC National Diplomas, International Baccalaureates and so on. This can easily result in a university rejecting or losing students who are smarter and have achieved more at school than other students whom the university happily accepts. It can also result in a university accepting students whose achievements at school should have shown that they are not up to the job. (When members of the DMU School of Computing have manually mapped entry qualifications to results, they have found the results very revealing, but doing it is very hard work.)

Of course, different university courses require different combinations of abilities, and the various academic qualifications and tests that applicants take provide indicators (of greater or lesser accuracy) of different abilities, so university admissions decisions ought to reflect the different demands of different subjects.

The situation is made more complicated by the fact that exam performance can be influenced by the quality of school education and other environmental factors, so that talented people from underprivileged backgrounds can underperform. University teachers and administrators believe strongly that they shouldn't reject better students in favour of worse on the basis of ethnicity, social class, sex, location, etc, but it isn't easy to make appropriate (rather than inappropriate) allowances for potential handicaps. And mature students can often do very well after underperforming at school, so judging their qualifications is different.

Some institutions including DMU have tried to reduce the workloads of university teachers by transferring most admissions decisions to administrators who make offers according to mechanical procedures using simple criteria. It should be possible to make this process more sophisticated.

The challenge in this project is to design and implement a system architecture that makes it easy to specify rules that cover a wide range of qualifications and tests, only some of which will be applicable to any given applicant, and that can combine them to produce decisions according to different criteria for different subjects.

Extensions

When intelligent systems are used in real-life applications, often the big challenge is maintaining them: correcting and updating the knowledge they hold as the situations they make decisions about change. How can you design your system to make it easy to alter and extend its knowledge base?

The system would benefit from a sophisticated user interface that makes it easy to use the system and edit its knowledge base.

Variant

You might like to develop a system to support the swift decision-making that needs to be done during Clearing, when applicants ring up and want almost immediate decisions.

Cross-Reference

I've also proposed a similar-looking project entitled Evaluating University Admissions Qualifications. The difference is that while this project is intended as an AI project, developing skills in designing and building rule-based reasoning systems, the Evaluating University Admissions Qualifications project is an information systems project where the scope for sophistication lies in the generation of statistical analyses of data and in handling complex databases. You might like to think of that as a system that could provide the evidence to support the rules used in the intelligent system proposed here.

Back to