Applying Trustworthy Assurance—Digital Mental Healthcare at UK Universities¶

The University Context¶

Prior to the onset of the Covid-19 pandemic, the suggestion that student mental health was in crisis across UK universities was already prominent.¹ Media attention intensified around 2017, for instance, in response to a cluster of high-profile suicides.² Since then, concern has only grown further, and focus has turned to whether the crisis has worsened due to the increased social isolation brought about by Covid-19 and remote learning.³

While media reports have been criticised for their simplistic focus on suicide figures as a metric for student wellbeing, research by The Office for Students has found that lengthy waiting times for counselling and a rise in help-seeking behaviour have both put an increased strain on services. All this has occurred during a period where the overall student population has grown.⁴

Research continues to point to significant challenges facing university mental health services given factors such as the fivefold increase in students disclosing mental health conditions between 2007 and 2017⁵ and the lack of capacity to address student concerns quickly. For instance, NUS research found that only one in six students received professional support within one week of reaching out.⁶

In this high-pressure context, systemic changes have been proposed. Key policy documents, from the University Mental Health Charter to Universities UK Step Change Framework and IPPR’s report on student mental health, have all called for a “whole university approach”, for better cohesion between university departments, and for a greater focus on positive wellbeing. In addition to structural shifts, these recommendations have increasingly referenced ‘digital interventions’ as a key tool within the student mental health offering.⁷

Digital interventions are frequently seen as a “natural step” for student mental health services as they expand beyond the traditional counselling model,⁸ while youth mental health is seen as a prime target for AI with the NHS AI strategy noting their first task with regard to mental wellbeing “is to look into children and young people’s mental health” using AI-driven solutions.⁹

Amid these policy recommendations for modernisation, digital mental healthcare offerings designed for the general population have proliferated with The Office for Students suggesting 43,000 wellness and medical apps are now available for smartphone use.¹⁰

It is no surprise, therefore, that these new tools have appealed to universities facing rising service demand alongside tight budgets, and that digital mental healthcare adoption has advanced at pace across UK universities.

One way in which the growth in digital mental healthcare has been amplified is through universities’ informal recommendations of external apps. For example, among the 24 Russel Group Universities in the UK, 15 recommend students try Headspace, eight point in the direction of Calm, and nine recommend SAM (Self-help App for the Mind), to name just a few of the apps listed on university webpages, typically under the heading of ‘self-help resources’.¹¹

Another way in which digital mental healthcare has infiltrated the university sector is through the formal procurement by universities of digital mental healthcare tools designed with the specific needs of student populations in mind. Kotouza et al. have mapped these shifts and found that, as of 2021, “more than half of UK universities” have a contract with at least one from SilverCloud and Togetherall while “Fika won over 35 UK university contracts between 2019 and 2020”.¹²

This rapid adoption of DMHTs may be seen as a perfect solution for the university sector, especially when the claims made by these digital service providers are taken at face value. For example, a co-founder of Fika has claimed that “mental fitness approaches like Fika’s present a new, far more sustainable solution to the natin’s mental health needs”,¹³ while Kooth describe their platform as a “safe and confidential space to share experiences and gain support”.¹⁴

However, it is important that universities are given the tools required to assess these claims themselves, and to do so using systematic, reliable, and well-validated procedures and standards. The University Mental Health Charter, for instance, emphasises that the “need for quality assurance extends to other interventions, such as the provision of digitally based services”. Further guidance is required for universities to be able to carry out such quality assurance, and in particular for university teams to effectively assess these digital tools on the basis of their ethical implications. Here, we present findings to inform universities in doing such due diligence.

Workshop Information¶

To assess current challenges to the ethical and trustworthy deployment of digital mental technologies in a university context, we ran a series of participatory engagements to seek the feedback of university students and mental health teams.

The findings presented below are based upon a series of participatory engagement events conducted by The Alan Turing Institute between January and March 2022. These consisted of semi-structured interviews with university administrators leading mental health teams at 10 universities across the UK (henceforth ‘administrators’) alongside research workshops attended by 25 students enrolled in undergraduate and postgraduate courses at UK universities (henceforth ‘students’).

Table 3.1—summary information about the two sets out workshops.

Analysis and Contextual Challenges¶

Key findings from these engagements conducted with university administrators and students are presented as follows. First, we explore contextual challenges to the ethical deployment of DMHTs at universities, as raised by students and administrators. Second, we present findings on how the implementation of trustworthy assurance to this sector may resolve challenges. In each case, distinct perspectives from administrators as compared to students will be highlighted and conclusions for the future of ethical digital mental healthcare procurement summarised.

As discussed in the Introduction, digital technologies raise a distinct set of ethical issues when deployed in a mental health context. For example, it is necessary to consider their impact on the therapeutic relationship and the privacy implications of using sensitive health data.¹⁵ Many of these ethical issues continue to be relevant at UK universities. However, the challenges posed by emerging DMHTs are specific to the context in which they are deployed. Therefore, this sub-project set out to determine HE sector-specific challenges to the trustworthy and ethical procurement and assurance of DMHTs.

Contextual challenges, ranging from obstacles to transparency and accessibility to external pressures and institutional constraints, are summarised to inform key conclusions for the HE sector.

Challenge 1: Duty of care: a legal or ethical goal?¶

To determine which digital technologies to procure and provide, universities must first have a clear understanding of the overarching goal of their service. Defining this top-level goal also constitutes the first stage in developing a trustworthy assurance case.

In a university context, defining this overarching goal introduces its own set of challenges. Universities are understood to have a duty of care to their students’ mental health. This has been defined by AMOSSHE—the Association of Managers Of Student Services in Higher Education—as the duty of the institution to “to deliver its educational and pastoral services to the standard of the ordinarily competent institution, and, in carrying out its services and functions, to act reasonably to protect the health, safety and welfare of its students.”¹⁶ Therefore, as part of this sub-project we explored whether 'duty of care' could serve as a top-level normative goal for a trustworthy assurance case—going beyond the SAFE-D principles.

All digital technologies must support the university in delivering upon this duty. But, to assess whether a particular procurement choice is in-keeping with a university’s responsibility, more detail is needed to translate duty of care into a fully specified goal.

For university administrators, our engagement found that legalistic understandings of duty of care were common as references to the avoidance of negligence and to ensuring institutional competence and compliance were frequent. This is unsurprising, and as one noted that, “if the worst outcome happens and you’re in a coroner’s court”, you must be able to show that due diligence has been done and procedure has been followed. Moreover, students also made frequent references to the importance of legal mechanisms in holding institutions to account.

However, despite these legalistic framings, it is clear that the law does not provide a complete guide to action. AMOSSHE’s account on duty of care clarifies these challenges, noting first that “student law is still evolving”, thus creating “difficulty in providing a legal definition of an institution’s duty of care”, and second, that “there is a balance between what the university should do as a legal minimum and what they could do based on a university’s perceived moral obligation”. These uncertainties can cause confusion for administrators.¹⁷

Ambiguities with regard to the nature and scope of a university’s duty of care can be observed when universities must balance tensions between the preservation of student autonomy and the management of risk through forced intervention. Administrators raise the difficulty of these decisions on whether to intervene, illustrating uncertainty around duty of care.

The everyday practices of a university mental health team must, therefore, go beyond the pursuit of compliance to grapple with ethical decisions on when specific interventions may or may not be appropriate. In short, legal understandings of duty of care must be supplemented by ethical values which guide the actions of teams in their selection of services.

For administrators, the ethical values raised as key to duty of care include: ‘inclusivity’, ‘putting the student first’, ‘consent’, ‘avoiding negative impact on the institution’, ‘best interest of every student’, ‘evidence-based decisions’, ‘benefit to all student groups’ and ‘value for money’. For students, key ethical values raised include: ‘transparency’, ‘inclusivity’, ‘anonymity’, ‘autonomy’, ‘privacy’ and ‘trust’.

While these ethical goals do show significant overlap, it is also clear to administrators that perspective is crucial to defining duty of care. First, different stakeholders define the goal of student mental health services differently. Second, perspectives on duty of care have shifted over time. Finally, the very introduction of digital services can influence who is seen to take responsibility for mental ill-health and so have an impact on how duty of care is understood.

Continued work is needed to understand if and how duty of care can be translated into a fully specified goal for a university mental health team and how this goal can inform specific actions during the procurement of DMHTs. A participatory approach, considering the views of all impacted stakeholders, will be essential to clarifying the nature and scope of this responsibility.

Challenge 2: Organisational constraints: defining roles and responsibilities¶

Ambiguities surrounding the relative roles and responsibilities of different teams and institutions to protect student mental health have also at times hampered effective decision-making with regard to digital mental healthcare. Descriptions of the decision-making rationale provided by administrators suggest a procurement process which is frequently reactive rather than proactive, with a tendency to stick to the status quo or shift tack dramatically during a crisis.

Reactive decisions are in part a consequence of the lack of clarity surrounding who is responsible for decisions on the procurement of DMHTs. First, there are significant ambiguities surrounding the relationship between university mental health services and the NHS. During engagements, administrators revealed that responsibility for procurement of digital services such as Togetherall resided in most instances with the university, but in certain cases with the NHS. In particular, administrators suggested that at times of crisis, where one institution was struggling to cope, the other had stepped in to fill a gap through purchasing a digital service. This dual responsibility can result in inconsistencies in who takes procurement decisions and in the long run could create a vacuum of responsibility if it is unclear where leadership lies.

In addition, within the university, shifts towards a ‘whole university approach’ have occurred and were described by administrators. However, while greater involvement of departments such as financial services, accommodation, and examinations is taking place, less thought has been given to where responsibility for digital solutions may lie within this matrix and what capacity building is needed to develop the technical skills required to review digital technologies, monitor student use, and horizon scan for new directions. While procurement processes at many universities remain thorough in many respects, this results at times in decisions taken by clinical or legal teams, without dedicated attention to technical and ethical questions. In describing their procurement teams’ top priorities when reviewing digital technologies, administrators raised both clinical evidence and value for money. Other concerns specific to data-driven technologies and their ethical implications did not feature as prominently, indicating potential gaps in data literacy.

Finally, as roles and responsibilities are clarified, a clear space for student involvement must be carved out. This was a priority raised by both students and administrators during engagements, but practical challenges must be addressed to involve all students and resolve possible tensions between student perspectives and clinical evidence. Additional challenges to student involvement in the procurement of digital technologies, such as the high entry costs sometimes associated with the evaluation of AI-driven tools, must also be addressed through education and collaboration between students and experts.

Challenge 3: External Pressures¶

Greater clarity on responsibility for digital technologies within university mental health teams will only go part of the way to resolving current challenges. Significant pressures placed on mental health teams are outside of their control. In particular, concerns were raised throughout interviews and workshops surrounding the potential privatisation of student mental health.

The relationships between universities and developers do suggest for-profit companies are having a significant influence on the procurement of DMHTs. The constant flood of digital offerings creates an environment where mental health teams struggle to ensure services are driven by student demand. Additionally, the nature of pitches received from developers often fail to provide the evidence or information necessary to evaluate services. Finally, the volume of products on the market has led to a dichotomy whereby administrators either spend significant time filtering through proposals or ignore them altogether.

These pressures which prevent services from being driven by student need are not only coming from private sector developers but from within the university sector itself. First, pressures from within an institution can lead to the adoption of a digital service on the basis of reputational concerns. Second, network pressures from comparisons with other universities can drive groupthink with regard to DMHTs.

Challenge 4: Algorithmic aversion: diverted resources and prioritising human services¶

In a university context, where resources are constrained and digital solutions have been described as offering value for money, it is unsurprising that fears have been raised over technologies replacing in-person services, despite this going against the desires of both administrators and students. Determining when digital solutions are not appropriate, desirable, or morally permissible, therefore, constitutes a key challenge for the higher education sector. Administrators note that despite their intention to supplement human services with digital, there have been times where more could be done to avoid the diversion of resources to digital technologies.

Appeals to the importance of prioritising in-person services are significant. However, further specificity is required to determine how this can be achieved in a system where resources are constrained, and administrators inevitably face tough choices on where to direct funds. A more detailed understanding of students’ and administrators’ motivations for algorithmic aversion is necessary, therefore, to determine when and how digital technologies should be delivered. For administrators, concerns over clinical efficacy and the management of risks on platforms, as well as a lack of demand from students, formed the primary motivations for algorithmic aversion. In contrast, students focused on concerns around the dehumanisation of mental health, the lack of empathy offered by technology and the potential exacerbation of social isolation by digital offerings. Further research into these and further motivations for students’ and administrators’ algorithmic aversion is essential if student mental health services are to effectively deploy digital technologies.

Taking these motivations into account can help to reveal a more complex picture where students and administrators oppose digital technologies for specific reasons and in specific circumstances. This detail is necessary to guide administrators in choosing when and how to deploy digital technologies, and when to stick with in-person services.

Challenge 5: Accessibility and fairness¶

In light of these varied motivations for aversion to digital solutions, it is clear that a more thorough understanding of different students’ needs is essential. During engagements, key concerns were raised surrounding accessibility and fairness in digital mental healthcare, as many participants proposed a more tailored approach was needed. While increased accessibility of mental health services was described by both students and administrators as a primary advantage of digital technologies, a one size fits all approach has meant that access to effective care has not been improved for all groups.

These limitations of digital technologies’ ability to improve accessibility need not be a reason against their deployment. In some instances, the ability to reach different groups, such as those who feel the stigma of reaching out in person, can be seen as a positive. However, to ensure accessibility is prioritised across the student population, a greater awareness of who is (and who is not) served by digital technologies is required, alongside an in-depth understanding of which student needs are currently unmet by in-person services. During engagements, some administrators suggested males were engaging more with digital technologies, some pointed to females, others to differences in engagement across academic disciplines and finally some noticed no patterns of engagement at all. Further research into this is needed for universities to move away from a one-size-fits all approach and design solutions which explicitly address the needs of all student groups.

For students, concerns about fairness and health equity were significant. First, concerns were raised around digital poverty and the importance of challenging assumptions frequently made about student populations (e.g., that all students have easy access to technology). Second, students expressed worries about biases within these technologies and how algorithmic bias can impact their experiences. Here, students advocated for specific consideration of minority student groups with the possible designation of ‘safe spaces’ on such platforms based on protected characteristics or allowing students to actively choose who they were speaking to in order to avoid potential bullying or harassment.

During discussions on the accessibility and fairness of digital technologies, proposals were made by students for accessibility to be understood in more subtle ways whereby digital solutions are not simply rolled out across the student population, but instead tailored to students with specific needs. It was also proposed that humans should mediate access to technology where possible so that a student can be matched with a service that helps them at the level they are ready for. Finally, empowerment of student choice was prioritised by participants in order to ensure students have control over the care they receive.

Challenge 6: Transparency and communication¶

Both students and administrators raised the importance of transparent and accessible communication about what services are available to students. Both groups also recognised that there is a lack of cut-through in current communication strategies leading to students being unaware of which services are on offer at their institutions.

However, among students, the need for more awareness on what services are available was not the first priority regarding transparency and communication. Rather than focus simply on advertising services to students through multiple channels, students emphasised the importance of universities communicating the limitations and personal costs associated with these technologies to them, such as the loss of privacy or limited clinical efficacy. Such transparency is also emphasised as important to facilitating informed consent.

Methodological Challenges¶

Overall, feedback from administrators and students suggests there is a long way to go to ensure responsible digital mental healthcare innovation and procurement across the UK HE sector. Progress can be made by following key recommendations set out here for universities. However, significant challenges will remain so long as changes do not take place elsewhere in the digital mental healthcare ecosystem. Systemic changes from developers, policy-makers, and university leadership teams are required to leverage the collective power required to foster a responsible innovation landscape.

The trustworthy assurance methodology provides one route through which such a culture of responsible innovation and transparent communication could be facilitated so that university stakeholders are given all necessary information to review new mental health tools offered to them. Due to the potential of this methodology in this sector we asked administrators to provide feedback on its deployment. Crucially, feedback indicated that this methodology would fit within current procurement practices, provide structure for the critical assessment of developers’ claims, and encourage more transparent communication among university staff about the rationale behind procurement decisions.

Challenge 7: Interpreting assurance cases¶

Trustworthy assurance cases (as set out in Chapter 2) can appear complex. This introduces two key challenges for administrators tasked with assessing whether a particular assurance case meets their requirements for the procurement of a new DMHT.

First, technical skills are required to assess such cases and ensure evidence matches with property claims. Second, assurance cases cannot be treated in isolation and must often be evaluated alongside clinical standards. It is important to avoid treating clinical efficacy and ethical standards as equivalent. The assurance exercise is not intended to provide a universal stamp of approval that ethical concerns have been addressed. Instead, it will require contextual interpretation on whether the evidence provided by a developer is sufficient for deployment in a specific context.

As a result, despite placing the greatest burden of responsibility on developers, this methodology does require administrators, as key decision-makers in the HE sector, to have capacity to effectively review assurance cases. Policy-makers must, therefore, take on responsibility for ensuring the necessary knowledge and organisational readiness to assess digital technologies is provided to universities. This could be offered through consultation with national bodies such as AMOSSHE, The Office for Students or Universities UK or through dedicating resources to capacity building within each higher education institution.

Challenge 8: Uptake by developers¶

To obtain necessary evidence and compile it into an assurance case requires time and resources on the side of developers. A key obstacle to the deployment of trustworthy assurance therefore rests upon the motivation of developers to do this due diligence. Currently, such motivation will be minimal as standards across the sector place few demands on developers.

Nevertheless, the collective purchasing power of universities across the UK is significant and can be leveraged in order to place further requirements on developers. This may be done through standards setting at the national level, for example through networks such as AMOSSHE. Additional pressures can also be placed on university networks by organisations such as the National Union of Students whose collective voice can influence procurement decisions.

In setting high standards for developers, there must also be efforts not to exclude smaller mental health providers from having the resources to produce assurance cases. For this reason, the burden placed on developers may be reduced by sharing best practices or building repositories of publicly accessible assurance cases and argument patterns which can be used as a starting point for developers as they embark on the ethical reflection process. These routes towards cross-sector best practices are discussed in greater depth in Chapter 5.

Challenge 9: Use as a communication tool¶

Finally, in a university context, concerns were raised regarding the utility of trustworthy assurance in communicating to students that due diligence has been done and ethical implications have been considered.

In conversation with the students, we discussed whether there was perceived value in administrators sharing assurance cases directly with the student body. Although there was no negative response, concerns were addressed about the complexity of assurance cases. Similar concerns were raised by administrators. In brief, stakeholders thought assurance cases represented too much information for effective communication, and that a summary would be preferable. Access to the full case could be made available for any students who were interested in order to facilitate both accessible communication and full transparency.