Dr Andrew Harvey
First published in The Australian on 16 May 2012.
Assumptions can be dangerous. In Britain, Deloitte offers a predictive analytics service that examines students by social background and demography to determine their likelihood of dropping out of their courses.
Universities can pay to know which types of students are likeliest to withdraw, and shape their admissions and support policies accordingly.
On first glance this is nothing new; all Australian universities would know the success rates of their identified equity groups, for example. Nevertheless, ethical questions are raised by using demographic data to make predictions.
Additional funds are already being provided to universities for every student enrolled from a low socioeconomic background. This money is being provided on the assumption that low SES students will need extra support at university, despite all the evidence showing they have similar rates of progression, retention and service use as the overall university cohort.
Federal funding reflects a predictive approach to equity, where risk factors are identified or assumed before admission.
The development of data capture systems such as the Victorian student number and the Commonwealth Higher Education Student Support Number will enable greater monitoring of progression from primary school to university. Longitudinal data will allow universities to discover which feeder schools are riskiest, which alternative entry pathways are effective and which prospective student groups are statistically likely to drop out.
While it seems logical for universities to analyse pre-entry data, there is a demographic overlay to this project that carries a risk of profiling.
There is only a small risk of school partnerships or pathways being changed to favour access for certain demographic groups. However, there is a real risk of students being labelled and stigmatised when they enter the university system.
Students who enrol via at-risk schools or schemes could be directed on enrolment to attend additional tutorials and academic support sessions on the basis of their admission pathway. They may be assigned an academic mentor and student buddy, and urged to complete multiple surveys on their transition to university life. This process portends a reification of identities, with support allocated by association rather than individual need.
Since many entry pathways are targeted to particular demographic groups, accusations of racial profiling, sexism and class snobbery may follow.
Under the guise of predictive analytics, under-represented students could automatically be stereotyped as risks for the university to manage.
A better method is to assume Rawlsian blindness at the point of entry and develop an effective early warning system. Some universities have already moved to establish a dashboard of student risk indicators, such as failure to log on to the email or learning management system, failure to submit initial assignments and failure to attend tutorials. These indicators can trigger a red flag that results in additional support being provided as required. No student is treated differently at the point of entry, with the early warning system being blind to demography, status, tertiary rank, school of origin, course and campus of enrolment.
The system is individualised and evidence-based and is not reliant on demographic and other assumptions to determine likely withdrawal.
There are technical barriers to developing an early warning system, but there is another factor that mitigates against its efficacy. Large numbers of students withdraw around orientation week, long before any warning system could detect risk indicators.
This reality means universities will continue to seek pre-enrolment risk signals, and the threat of inadvertent profiling will remain. Without careful treatment of data, universities may exacerbate stereotypes at the very time they are trying to eliminate them.
Andrew Harvey is head of the Access and Achievement Research Unit at La Trobe University, Melbourne.