The New Transcript and Predictive Analytics
Only a Matter of Time?
As interest in alignment between education and industry increases, higher education institutions are looking for new ways to signal their students’ industry-relevant skills and experiences to employers in ways that are meaningful and practical.
A promising example is the “new transcript” that a number of US colleges are developing. The new transcript includes information that is more readily translated into job skills than traditional transcript data, such as specific course learning outcomes and hours spent on extracurricular activities and internships. Furthermore, the new transcripts are digital and easy to share with employers, whose interests are key to this process. And, as Georgetown University’s Anthony Carnevale has pointed out, the information could be used to create accountability metrics for higher education. For instance, digitized aggregate transcript data allows colleges to assess which combinations of learning experiences lead to the best-paying jobs.
As digitized transcripts become more sophisticated and comprehensive they could go beyond shedding light on college-to-job pathways and actually become a job placement tool. Just as a student’s administrative records and learning management system data can be used to predict their chances of say, timely college degree completion, it is easy to imagine applying educational data mining and predictive analytic techniques to new transcript data to predict a graduate’s chances of success at a particular type of job. Not only could colleges use this information to provide proactive and targeted career guidance and support, but employers could use it to make hiring decisions and avoid the controversies of nebulous “fit” hiring.
Of course, this would first require years of new transcript and labor market data, as well as standard definitions for job categories and job success metrics. But it could move very quickly as employers would have a strong incentive to help advance such models, students are increasingly comfortable sharing their digital education data publicly, and models can be tested on existing online alternatives such as public LinkedIn profiles.
There are risks to this progression, however. For instance predictive models and their underlying datasets can reproduce or create bias, leading to interventions that can systematically disadvantage particular groups of students. Moreover, there could be undesirable effects on college curriculum and assessment. Some student skills and learning outcomes that are harder to spell out and track, including those that are more subjectively evaluated, not vocation-oriented, or obtained outside digital environments, may become undervalued. Because of this, some academics worry that the quantification of learning into marketable outcomes will misrepresent students’ college experiences and propel institutions to focus too narrowly on career readiness. More importantly, employers could use new transcript data to make hiring decisions without a full understanding of the research behind the predictive models and its implications, and without oversight from governing bodies that ensure the ethical and responsible use of student data.
It is difficult to envision all the advantages and pitfalls of such use of student data, especially when considering the possibility of its use by employers. While the new transcript is still in progress and its relationship with employers is not yet defined, its proponents might wish to consider these possibilities and draw on lessons learned from the fast growth and deployment of predictive analytics in educational settings to inform the process.