Ethics and Learning Analytics: Charting the (Un)Charted

As the field of learning analytics matures, and discourses surrounding the scope, definition, challenges, and opportunities of learning analytics become more nuanced, there is benefit both in reviewing how far we have come in considering associated ethical issues and in looking ahead. This chapter provides an overview of how our own thinking has developed and maps our journey against broader developments in the field. Against a backdrop of technological advances and increasing concerns around pervasive surveillance and the role and unintended consequences of algorithms, the development of research in learning analytics as an ethical and moral practice provides a rich picture of fears and realities. More importantly, we begin to see ethics and privacy as crucial enablers within learning analytics. The chapter briefly locates ethics in learning analytics in the broader context of the forces shaping higher education and the roles of data and evidence before tracking our personal research journey, highlighting current work in the field, and concluding by mapping future issues for consideration.

pointed to the increasing importance of learning analytics as an emerging technology, which has since developed from a mid-range technology or linkages between learning analytics and the more are also important distinctions regarding, inter alia, automation, aims, origins, techniques, and methods and practice (see van Barneveld, Arnold, & Campbell, 2012), so too thinking around ethical issues has slowly established one of the earliest frameworks developed with a focus on ethics in learning analytics. Since then, increasing surveillance and the (un)warranted colto an atmosphere of uncertainty among potential ben-ethical implications of learning analytics, before We then consider recent developments and conclude broader and more critical engagement.
There is some consensus that the future of learning will be digital, distributed, and data-driven such that sophisticated data collection and; (c) advanced machine ethical implications of learning analytics were initially has come a long way and is increasingly foregrounded be said for the potential economic students and the institution) of more successful harvesting, we should not ignore the possibilities of -Ethical implications around the collection, analysis, from the collection, analysis, and use of student data will depend on the interests and perceptions of the particular stakeholder. In this chapter, we hope to provide insight into the different positionalities, claims, and interests of primarily students and institutions.
Although now becoming more established, ethics and the need to question how student data is used and under what conditions was very much a marginal majority of sessions at this conference remained focused on developmental work. There was some mention of stakeholder perceptions of the ways in surveyed stakeholder considerations were largely focused on privacy and not considered particularly 2012) touched upon the need to consider the impacts of all stakeholders on students' learning journeys in order to increase the success of students' learning.
to map the challenges and opportunities, but also impact on student success and retention. At the same a number of assumed relevant ethical issues from different stakeholder perspectives.
ing institutional policy frameworks that set out the purposes for how data would be used and protected rapidly. In general, policies relating to institutional use of student data had not kept pace, nor taken account focusing mainly on data governance, data security, Taking a sociocritical perspective on the use of learnthe ethical use of learning analytics. A range of ethical issues was grouped within three broad, overlapping categories, namely: • The location and interpretation of data These principles offer a useful starting position, but ought sensibly to be supported by consideration of a number of practical considerations, such as the development of a thorough understanding of who and opting out; issues around vulnerability and harm (e.g., inadvertent labelling); systems of redress (for both student and institution); data collection, analyses,

ESTABLISHING ETHICAL PRINCI-PLES: HOW FAR HAVE WE COME?
access, and storage (e.g., security issues and avoiding perpetuation of bias); and governance and resource allocation (including clarity around the key drivers straints, and the conditions that must be met).
This latter aspect of resource allocation was carefully learning analytics offers theoretical opportunities for HEIs (higher education institutions) to proactively identify and support students at risk of failing or dropare (increasingly) limited. The challenge then is where best to direct support resources and on what basis that decision is made. The concept of educational triage as a means of directing support toward students most balance between respecting student autonomy and, at the same time, ensuring the long-term sustainalways act in the student's best interest); the need for distributive justice (understanding that demographic characteristics have and do impact support provided and address this).
An increasing awareness of learning analytics as a means of doing something to the student without that student necessarily knowing triggered further Slade, 2014b). The resulting discussion challenged assumptions around learning analytics as a producer of accurate, objective, fully complete pictures of student learning, and also reviewed the potentially unequal relationship between institution and student. elements that could form a basis for a student-centred learning analytics: 1. essential in delivering effective and appropriate teaching and learning, but students should be able to make informed opt in/out decisions 2. Students should have full(er) knowledge of which data is collected and how it is used 3. Students should ensure that their personal data records are complete and up to date 4. The surveillance of activities and the harvesting of data must not harm student progress 5. Algorithmic output should be subject to (potential) human review, and corrected if needed of students, and algorithms should be frequently reviewed and validated students as active agents in the use of their own data Open University (OU; 2014) policy on the ethical use of student data for learning analytics. As part of the stakeholder consultation, a representative group of 50 which data is used to support students in completing their study goals over a three-week period. A study of was already actively collected and used, and they raised a number of concerns. The major concern related to the potential to actively consent (or not), with a opt out. This direct involvement of student voices in shaping a policy dealing with the ethics of learning analytics offered unique insight into the ways in which students regard their data -as a valuable entity to be carefully protected and even more carefully applied.
not be fully representative of the total population, the In response to this growing awareness of student conassumptions and understanding of issues surroundby both the apparent ease with which the public appear to share the detail of their lives and by our largely paternalistic institutional cultures. The study simple choice to allow students to opt-in or opt-out of having their data tracked. As a foundation for the discussion, the terms and conditions of three massive open online course (MOOC) providers were reviewed to establish information given to users regarding the how HEIs can move toward an approach that engages and more fully informs students of the implications of learning analytics on their personal data. A similar paper challenged the tendency for many HEIs to adopt the rapid growth in the deployment of learning analytics, few HEIs have regulatory frameworks in place and/or are fully transparent regarding the scope of asymmetrical information and power relations with agency in learning analytics. The aim was to consider ways in which student vulnerability may be addressed, increasing student agency, and empowering them as active participants in learning analytics -moving from It is broadly accepted that the increasing value of change value has overtaken our legal and traditional costs of capitalism's new data relations for our very such, there have been attempts in different geopolitical higher education in the United States, Australia, and contributions for 1) quality assurance and quality improvement; 2) boosting retention rates; 3) assessing and acting upon differential outcomes among the student population; and 4) the development and introduction of adaptive learning. The report acknowledges the many opportunities, but also highlights threats such concerns in learning analytics is that of The Open delimiting the nature and scope of data collected and not be collected and used for learning analytics. The 1. Learning analytics is an ethical practice that such as open entry to undergraduate level study.
2. The OU has a responsibility to all stakeholders to 3. principle furthermore warns against stereotyping students and acknowledges those individuals who makes it clear that members of staff may not have access to the full data set, which can seriously 4. The purpose and the boundaries regarding the and visible.
5. The University is transparent regarding data collection, and will provide students with the opportunity to update their own data at regular intervals.
Students should be engaged as active agents in the implementation of learning analytics (e.g., interventions).
7. Modelling and interventions based on analysis of data should be sound and free from bias.
8. Adoption of learning analytics within the OU re-cal implications in the collection, analysis, and use of student data, this policy and its principles attempted the purposes to which some or all of their data may be used for learning analytics and provides consent. Informed consent applies at the point of reservation University, 2014, p. 3). The policy does not address the possibility of students who prefer to opt out of the collection, analysis, and use of their data (as discussed In a recent overview of learning analytics practices in sector has not been ready for such a conversation previously, although it is argued that as institutions are maturing, ethical considerations take on a heightened -

RECENT DEVELOPMENTS IN ETHICAL FRAMEWORKS
of the discipline with institutions, practitioners and the real potential for abuse/misuse and discrimination (p. 588). Of particular importance is the commitment lytics practices that use data sources: (a) not directly related to learning and teaching; and/or (b) where which the data in question was collected; its use can continue under the following conditions: those who are the subject of measurement. Where informed consent means that: (a) clear and accurate information is provided about what data is or may be collected, why and how it is collected, how it is stored and how it is used; and (b) agreement is freely given to the practice(s) described. (p. 590) The above principles should be read in conjunction with two remaining principles regarding how collected data should be used to enhance teaching and of transparency and informed participation. For a full discussion, see Welsh and McKinney (2015). of ethics, privacy, and respective legal frameworks, and highlight challenges such as the real possibility relationship between data gatherer and data object, issues of ownership, anonymity and data security, privacy and data identity, as well as transparency and that learning analytics proceeds in an acceptable and legal, and logistical issues in learning analytics with an overview of how a range of stakeholders, such as senior management, the analytics committee, data scientists, educational researchers, IT, and students are impacted and have responsibility in learning analytics. The draft covers a wide range of issues including, inter alia, consent; identity; potential impacts of opting out; the asymmetrical relationship between the institution and students; (boundaries around) the permissible uses of student data; transparency; autonomy, amongst others. See Sclater (2015) for a full list of ethical concerns. al. (2015) consider the implications of the Law for the lytics. These include the need for permission (and the responsibility arising from receiving consent) and the implications of the consensual agreement between a service provider and recipient that the provider may use any personal information needed for the provision of the service. The law distinguishes between essential (2015) take a contentious view that, given that learning analytics is seen as an emerging practice, it may safely agreement between the institution and students. The authors suggest that these four principles should guide learning analytics: • and purpose for which it was provided • Subsequent use of such data should be reconcilable • for transparency, student consultation, and buy in • the governance of their data, including the following: • Easy access to collected information • The right to correct wrong information (or interpretations arising from it) • The right to remove irrelevant information implications for algorithmic decision making and the take responsibility for and have oversight of algorithmic decision making. Algorithms may, at most, signal particular behaviour for the attention of faculty or support staff. Further, students have a right to appeal decisions made based on analyses of their personal data. In cases where HEIs subcontract to software developers, the final responsibility and oversight remains securely with the institution and cannot be delegated (see Engelfriet et al., 2015).
It falls outside the scope of this chapter to map current ities and practicalities at the intersections between SOME FUTURE CONSIDERATIONS student data and advances in technology and methods of analysis. We would like to conclude, however, with some pointers for future consideration. ensure effective, appropriate, cost-effective learning there is broad agreement that institutions have a right to collect and use student information. However, there is no easily agreed upon position around consent, that is, in allowing students to opt out of the collection, analysis, and use of their data. Student positions logical or rational. The often-implicit calculation of in O'Brien. 2010).
Center for Fair and Open Testing in the US who encouraged students to refuse to take government-manbetween students' concerns, their right to opt-out, and the implications for the mandate of higher education to use student data to make interventions at an individual level. Central to this issue is the question of of the ethics around the collection, analysis, and use of student data (whether in learning analytics or in testing claims and vested interests. such as the increasing and persisting concerns about research ethics considerations with the use of online that many the researchers go beyond the Belmont principles (with the main emphasis on ensuring that outcomes outweigh potential harms caused by the ticipants, (2) ethical deliberation with colleagues, and There is also increasing concern balancing optimism addresses concerns regarding the potential harm not implemented with care, they can also perpetuate, It makes a number of suggestions relating to investment in research into the mitigation of algorithmic discrimination, encouraging the development and use of robust and transparent algorithms, algorithmic the roles of the government and private sector in setting codes of practice around data use. and use of personal data; the scope and nature of intrusion; the quality of the data and the automation of the decisions relating to the collected data; the risk of negative unintended consequences; whether the data objects agreed to the collection and analysis; the nature and scope of the oversight; and the security of the collected data. The framework also proposes a the risks to privacy and negative unintended conthe opinions of the data objects/public regarding the cation, and the increasing blurring of the boundaries between broader developments in data and neuroscience, we need a critical approach to considering the biopolitical strategy focused on the evaluation and management of the corporeal, emotional and embrained lives of to consider the basis and scope of authority of educaauthority to produce systems of knowledge about analytics in future will be essentially based on and driven by algorithms and machine learning and we maintain, or even reshape visions of the social world, regulatory frameworks will be essential elements in the frameworks ensuring ethical learning analytics While this chapter maps the progress in considering the ethical implications of the collection, analysis, and use of student data, it is clear that the potential for harm will not be addressed without further consideration of institutional processes to ensure accountability and transparency. As Willis, Slade, falls outside the processes and oversight provided by institutional review boards (IRBs). It is not clear at this stage by whom and how the ethical implications of learning analytics will be assured.
Since the emergence of learning analytics in 2011, nuanced in increasingly considering the fears and realities of ethical implications in the collection, analysis, and use of student data. In this chapter, we provide an overview of how our own thinking has developed a backdrop of technological advances and increasing concerns around pervasive surveillance, and a growing consensus that the future of higher education will be digital, distributed, and data-driven, this chapter maps how far the discourses surrounding the ethical implications of learning analytics have come, as well as some of the future considerations.
Each of the frameworks, code of practices, and conceptual mappings of the ethical implications in learning analytics discussed adds a further layer and a richer understanding of how we may move toward ness and appropriateness of teaching, learning, and student support strategies in economically viable and ethical ways. The practical implementation of that understanding remains largely incomplete, but still wholly pertinent.
We would like to acknowledge the comments, critical inputs, and support received from the editorial team and specially the reviewers of this chapter.