In March 2017, Manuela Ekowo and Iris Palmer co-authored a report for New America that offered five guiding practices for the ethical use of predictive analytics in higher education. This kind of work is really important. It acknowledges that, to the extent that analytics in higher education is meant to have an impact on human behavior, it it is a fundamentally ethical enterprise.
Work like the recent New America report is not merely about educational data science. It is an important facet of educational data science itself.
Are we doing ethics well?
But ethics is hard. Ethics is not about generating a list of commandments. It is not about cataloging common opinion. It is about carefully establishing a set of principles on the basis of which it becomes possible to create a coherent system of knowledge and make consistent judgements in specific situations.
Unfortunately, most work on the ethics of analytics in higher education lacks this kind of rigor. Instead, ethical frameworks are the result of a process of pooling opinions in such a way as to strike a balance between the needs of a large number of stakeholders including students, institutions, the economy, the law, and public opinion. To call this approach ethics is to confuse the good with the expedient.
Where should we begin?
An ethical system worthy of the name needs to begin with a strong conception of the Good. Whether stated or implied, the most common paradigm is essentially utilitarian, concerned with maximizing benefit for the greatest number of people. The problem with this approach, however, is that it can only ever concern perceived benefit. People are famously bad at knowing what is good for them.
A benefit of this utilitarian approach, of course, is that it allows us to avoid huge epistemological and metaphysical minefields. In the absence of true knowledge of the good, we can lean on the wisdom of crowds. By pooling information about perceived utility, so the theory goes, we can approximate something like the good, or at least achieve enough consensus to mitigate conflict as much as possible.
But what if we were more audacious? What if our starting point was not the pragmatic desire to reduce conflict, but rather an interest in fostering the fullest expression of our potential as humans? As it specifically pertains to the domain of educational data analytics, what if we abandoned that instrumental view of student success as degree completion? What if we began with the question of what it means to be human, and wrestled with the ways in which the role of ‘student’ is compatible and incompatible with that humanity?
Humane data ethics in action
Let’s consider one example of how taking human nature seriously affects how we think about analytics technologies. As the Italian humanist Pier Paolo Vergerio observed, all education is auto-didactic. When we think about teaching and learning, the teacher has zero ability to confer knowledge. It is always the learner’s task to acquire it. True, it is possible to train humans just as we can train all manner of other creatures (operant and classical forms of conditioning are incredibly powerful). but this is not education. Education is a uniquely human capability whereby we acquire knowledge (with the aim of living life in accord with the Good). Teachers do not educate. Teachers do not ‘teach.’ Rather, it is the goal of the teacher to establish the context in which the student might become actively engaged as learners.
What does this mean for Education? Viewed from this perspective, it is incumbent on us as educators to create contexts that bring students to an awareness of themselves as learners in the fullest sense of the word. It is crucial that we develop technologies that highlight the student’s role as autodidact. Our technologies need to help bring students to self-knowledge at the same time as they create robust contexts for knowledge acquisition (in addition to providing opportunities for exploration, discovery, experimentation, imagination and other humane attributes).
It is in large part this humanistic perspective that has informed my excitement about student-facing dashboards. As folks like John Fritz have talked about, one of the great things about putting data in the hands of students is that it furthers institutional goals like graduation and retention as a function of promoting personal responsibility and self-regulated learning. In other words, by using analytics first and foremost with an interest in helping students to understand and embrace themselves as learners in the fullest sense of the term, we cultivate virtues that translate into degree completion, but also career success and life satisfaction.
In my opinion, analytics (predictive or otherwise) are most powerful when employed with a view to maximizing self-knowledge and the fullest expression of human capability rather than as way to constrain human behavior to achieve institutional goals. I am confident that such a virtuous and humanistic approach to educational data analytics will also lead to institutional gains (as indeed we have seen at places like Georgia State University), but worry that where values and technologies are not aligned, both human nature and institutional outcomes are bound to suffer.