Ethics in General
What Ethics Is
Ethics is one of those subjects where each of us feels we have an understanding of what it is and the role it plays in life, but where there is a history of deep and often contentious discussion. When one person says to another “this is unethical” it may appear that there is a common understanding between the two but on further examination there may be a great deal of difference.
Ethics, at first glance, appears to be about ‘right’ and ‘wrong’, perhaps as discovered (Pojman, 1990), perhaps as invented (Mackie, 1983). The nature of right and wrong might be found in biology, rights, fairness, religion, or any number of other sources, depending on who is asked. Or instead, ethics may be based in virtue and character, as described by Aristotle (2014) in ancient Greece. Either way, ethics is generally thought of as speaking to what actions we ‘should’ or ‘ought’ to take (or ‘should not’ or ‘ought not’ take).
What Ethics is Not
These differing perspectives on ethics will be examined more closely below, after a discussion of learning analytics, so that we may regard them in that context. But for now, it is worth taking note of what ethics is not (after Velasquez, et.al., 2009):
- While feelings may offer a basis for ethical choices (as we shall see below), ethics is not the same as feeling. So, for example, we would not say that the word ‘unethical’ means the same as the word ‘repugnant’ or ‘offensive’. These words describe our reactions to an action, but more needs to be said before saying that the action is unethical.
- While religion may offer a basis for ethical choices, ethics is not the same as religion. For one thing, it is arguable that a person could be ethical without God (Nielson, 1973). Conversely, it is arguably possible to base unethical actions in religion, perhaps, say, when doctors place their own religious beliefs above their patients’ interests (Dickens, 2009). Moreover, it is also arguable that the domain of ethics extends beyond religion (one wonders, for example, what God would have to say about copyright law).
- While law may help us decide whether an action is right or wrong, the concepts of ‘ethical’ and ‘legal’ are not the same (nor, arguable, should they be). It’s not simply that law can become corrupt (though it certainly can), but also that the law and ethics serve different purposes and apply to different domains.
- While cultural norms may influence our ethical decisions, it is arguable that different cultures define ethics differently, taking for example, a compliance-oriented perspective as opposed to a value-oriented perspective (Eisenbeiß and Brodbeck, 2014). And while some cultural practices may appear obviously wrong (Velasquez, et.al. (2009) mention slavery) other differences are much more subtle.
- While science may help us make ethical choices, science does not define right and wrong. Indeed, some argue that ‘ought’ does not follow from ‘is’, that is, we cannot infer to the way things should be from the way things are, while others disagree (Hudson, 1969). So it is not clear we can base ethics in physics, biology, or any of the other sciences.
What we should take away from this list is that there is not to be had a quick reduction of ethics to some other discipline. Ethics may involve any or all of these disciplines, each playing a role in informing our ethical choices, but none dictating them.
A Framework for Making Decisions
In the context of learning analytics, ethics might best be viewed as providing a framework for making decisions. This follows from the idea of analytics and related technology posing dilemmas requiring practitioners to make the right choice (whatever that may be). It also follows from a perspective that has, as Neil Selwyn (2019) says, “framed the use of data in education in what are “always intensely political and normative decisions” (Shelton, 2017, p. 24).
At the same time, while it may be that learning analytics is as Selwyn says “inherently political” but it does not follow that all political aspects of analytics are also ethical aspects. So while we may, as he says, discuss analytics “in terms of power, control, and disparities in access and outcomes,” it does not follow that all such discussions are discussions of ethics. There is a difference, I think it could be argued, between an outcome that is undesired, and an outcome that is ethically wrong.
To the extent that analytics includes artificial intelligence (AI) (and I will use the two terms broadly and loosely throughout this discussion) analytics includes some sense of rationality and decision-making. As Zawacki-Richter, et. al. (2019:4) write, “the concept of rational agents is central to AI… we are therefore dealing here with GOFAI (‘good old-fashioned AI’, a term coined by the philosopher John Haugeland,1985) in higher education–in the sense of agents and in-formation systems that act as if they were intelligent. Just as the concept of rational agents is central, arguably, to ethics.
What Are Learning Analytics?
Analytics is thought of generally as “the science of examining data to draw conclusions and, when used in decision making, to present paths or courses of action.” (Picciano, 2012). This includes not only the collection of the data but also the methods of preparation and examination employed, and the application of the data in decision-making. Thus the term ‘analytics’ can be thought of as the overall process of “developing actionable insights through problem definition and the application of statistical models and analysis against existing and/or simulated future data.” (Cooper, 2012)
The focus of this paper is the use of analytics as applied to learning and education (typically called ‘learning analytics’). Learning analytics is typically defined in terms of its objective, which is to improve the chance of student success (Gasevic, Dawson & Siemens, 2015). Accordingly, when founding the Society for Learning Analytics (SoLAR) George Siemens defined learning analytics as “the measurement, collection, analysis and reporting of data about learners and their contexts, for purposes of understanding and optimizing learning and the environments in which it occurs.” (Siemens, 2012)
Learning and Education
The focus of learning analytics is often described from the institutional perspective. For example, as Slade and Tait (2019) write, “The UK-based Higher Education Academy states that learning analytics offers the potential to provide educators with quantitative intelligence to make informed decisions about student learning… as well as to inform pedagogy, allocate resources and inform institutional strategy (Rienties, et.al. 2016).”
What we mean by ‘learning and education’ might be very different depending on who is being asked. For example, in one study, “senior managers were most interested in using (learning analytics) to improve institutional performance, whereas teaching staff to reform curriculum and improve student support, and students to receive more personalised education tailored to meet their needs.” (Tsai, et.al., 2018)
A Narrow Definition?
Some argue for a narrow definition of ‘learning analytics’. The term ‘academic analytics’ (Goldstein, 2005; Campbell, et.al., 2007; Long & Siemens, 2011) has been used to distinguish between technologies used by higher education institutions “to support operational and financial decision making”, as opposed to ‘learning analytics’, defined as technologies “focused toward instruction, curriculum, and learning support with the objective of achieving specific learning goals.” (Van Barneveld, et al., 2012). More recently, Zeide (2019) distinguished between institutional, student support and instructional applications.
Some argue for a narrow definition based on outcome. Luckin, et.al. (2016), quoting John Self (1998) suggest that “at the heart of AIEd is the scientific goal to ‘make computationally precise and explicit forms of educational, psychological and social knowledge which are often left implicit,’” or in other words, “to open up what is sometimes called the ‘black box of learning,’ giving us deeper, and more fine-grained understandings of how learning actually happens (for example, how it is influenced by the learner’s socio-economic and physical context, or by technology).” That might be one idealized objective of AI and analytics, though arguably Self’s 1999 perspective no longer describes contemporary learning theory, and in any case, as well shall see, there are much wider applications and benefits.
Another approach is to identify the different areas in which analytics is used. For example, the UC Berkeley Human Rights Center Research Team (2019) divides the domain into three categories: “AI tools fall into three categories: learner-facing, teacher-facing and system-facing.” However it quickly becomes apparent that the same tool will most probably have multiple faces; a learning management system, for example, may support student work, while analyzing and reporting progress to the teacher, while updating system-wide analytics.
A Broad Definition
From the perspective of ethics in analytics, it may be wisest to adopt a broad definition of ‘learning analytics’. After all, as Griffiths et.al. (2016) argue, the “Jisc Code of Practice for Learning Analytics uses a wider definition of using data about students and their activities ‘to help institutions understand and improve educational processes, and provide better support to learners’ (Sclater 2014b). In this case, we may wonder what the scope of ‘educational process’ is. Could it, for example, include recruitment and marketing? Is there anything that an educational institution does that is not an ‘educational process’?”
A wider definition not only avoids the difficulties of establishing a more narrow definition, but also ensures we do not disregard potential ethical implications simply because the practice is ‘outside the scope of learning analytics’. Instead, to provide a sense of what we mean by ‘learning analytics’, we will consider a wide range of applications of analytics in a learning or educational context, and then consider the ethical issues that arise from such applications.
Analytics and Artificial Intelligence
Arguing for a broader definition of analytics necessarily leads us to consider including artificial intelligence (AI) in the conversation. Generally, “Artificial intelligence (AI) systems are software (and possibly also hardware) systems designed by humans that, given a complex goal, act in the physical or digital dimension by perceiving their environment through data acquisition, interpreting the collected structured or unstructured data, reasoning on the knowledge, or processing the information, derived from this data and deciding the best action(s) to take to achieve the given goal.” (European Commission’s High-Level Expert Group on Artificial Intelligence, 2019: 36)
Indeed, through this paper, the terms will often be used interchangeably, not because they are thought to be the same, but because in the context of this paper, many of the same things could be said about both, especially in the domain of learning and education. However you define the terms, artificial intelligence plays a significant role in analytics, and vice versa, so we will treat them together as one thing (Adobe Experience Cloud Team, 2018). If a distinction is necessary during the course of the discussion, we will apply it.
It should be noted that ‘Artificial intelligence’ is generally defined more broadly than ‘analytics’. For example, AI systems can either use symbolic rules or learn a numeric model, and they can also adapt their behaviour by analysing how the environment is affected by their previous actions” (European Commission’s High-Level Expert Group on Artificial Intelligence, 2019: 36). For the most part, the AI and analytics under consideration in this paper are not based on symbolic rules, as the field has mostly turned away from such systems.