Integrated Care Systems as an Arena for the Emergence of New Forms of Epistemic Injustice

Epistemic injustice has rapidly become a powerful tool for analysis of otherwise hidden social harms. Yet empirical research into how resistance to knowing and understanding can be generated and replicated in social programmes is limited. We have identified a range of subtle and not-so-subtle inflections of epistemic injustice as they play out in an intervention for people with chronic depression in receipt of disability benefits. This article describes the different ‘species’ of epistemic injustice observed and reveals how these are unintentionally produced at frontline, management, commissioning and policy levels. Most notably, there remains a privileging of clinical knowledge over other forms of knowledge, producing a ‘pathocentric epistemic complex’. This, combined with the failure of different agencies with competing ideologies to adequately understand each other, and a vicious policy context, added to the injustices already faced by people with mental health issues, generating multiple harms. This has important implications for a range of integrated care and welfare interventions – not least by drawing attention to their unintended potential for replicating epistemic injustice as an institutionalised complex. Careful evaluation and design of such programmes, applying the philosophical and epistemic resources illustrated here, can help mitigate this outcome. Further, by raising awareness of epistemic injustice among programme participants, we can generate epistemic structures that secure programme integrity locally, and promote better policy.


Introduction
The concept, epistemic injustice, is at once simple to grasp yet imbued with the potential to take multiple hidden forms, twisting in response to the diverse contexts in which it often surfaces silently. The act of treating someone unfairly "in their capacity as a knower" (Fricker 2007: 1), and the distinctions of hermeneutical and testimonial injustice, provide a fruitful conceptual territory that continues to be explored and developed in theoretical and empirical realms. The active debate in this relatively young field has expanded on those initial distinctions to include notions around contributory injustice (Dotson 2012: 31-32), pathocentric epistemic injustice (Kidd and Carel 2019) and institutional epistemic injustice (Fricker 2007;Anderson 2012;O'Donovan and Madden 2018), maintained by epistemic communities (McHugh 2017). Some scholars have warned about perpetuating oppression through defining epistemic injustice in ways that could exclude important contributions to the discussion, and argue that the concept should be considered open (Dotson 2012;Pohlhaus Jr. 2017). It is therefore important to try to describe 'types' in context, understand their roles, causes and effects, and recognise how they manifest differently at different points in a system. Our evaluation of an integrated health and welfare programme in the UK provides a useful meso-level stage for catching glimpses of emergent forms.
Research on epistemic injustice has carved out a specific niche in relation to illness and mental illness in particular.
People experiencing mental distress are particularly vulnerable to epistemic injustices as a consequence of deeply embedded social stigma resulting in a priori assumptions of irrationality and unreliability such that their knowledge is often discounted or downgraded. (Newbigging and Ridley 2018: 36).
However, much of the literature on epistemic injustice and mental illness tends to focus on the pathologization and epistemic silencing of patients by health professionals, belying their stated commitment to person-centred care (Lakeman 2010;Sanati and Kyratsous 2015;Leblanc and Kinsella 2016;Crichton et al. 2017;Jackson 2017;Kyratsous and Sanati 2017;Scrutton 2017;Kurs and Grinshpoon 2018;Newbigging and Ridley 2018). Less attention is paid to identity prejudice between professional groups, or the ways in which the systems and institutions designed to support mental health can produce epistemic injustices that damage outcomes for patients in more insidious ways.
…while Fricker's focus on individual epistemic virtue is important, we also need to consider what epistemic justice as a virtue of social systems would require. (Anderson 2012: 163). Consider Fricker's paradigmatic case of epistemic injustice, the (1975) case of Carmita Wood, which explores how social understanding is mediated by inter-disciplinary interactions and professional discourse to arrive at expert judgements (Fricker 2007). Here, Fricker uses an underlying feminist analysis to shine a spotlight on the injustice that Wood had suffered in being subjected to unwanted sexual interest by a former boss, and finding herself ashamed and embarrassed to speak about this subsequently in front of a claims inspector for employment benefits. Her claim for benefits was denied. Our starting point is the fact that Carmita's claim today would in all likelihood be granted on the grounds of her suffering depression without a need to justify the reasons why (Viola and Moncrieff 2016; Department for Work and Pensions 2019). Such decisions involve blended domains of knowledge and inter-professional relationships between agencies. Our contemporary context reflects this progress but reveals new hindrances. For example: the widespread adoption of 'person-centred care', especially in mental health, highlights the value of client knowledge. 1 But this can still be undermined by bureaucratic and inter-professional disparity, access to knowledge, and power relations based on whose knowledge is considered authoritative.
This article focuses on the multiple inflections of epistemic injustice produced by agents and systems in 'Working Well', an integrated mental health and employment programme. Instead of focusing on the dynamics between professionals and service users, our study examined the relationships between experts in social care and welfare, mental health professionals, their managers, commissioners and policymakers, to identify several different forms in which epistemic injustice played out: 1) prior assumptions combined with experiences on the programme led to identity prejudice between practitioners; 2) communications software and bureaucratic processes unfairly filtered out some valuable forms of knowledge in conversations around clients; 3) ideological and political disagreement led to managers and commissioners failing to understand each other's positions; and 4) this was underpinned by a vicious welfare policy that, through the use of sanctions, undermined clients' trust in the system and consequently the programme. We argue that these injustices formed a complex that worsened outcomes for mental health service users who already face injustices based around social stigma.
Identifying dynamics that participants may already be painfully aware of and describing these through the somewhat abstract lens of epistemic injustice to create a new layer of understanding ironically risks generating an epistemic hierarchy. That is not our intention. This article highlights the multitude of potential injustices operating throughout the system. An alertness to this may inform future programme design. Our wider project, the ERC-funded 'Knowledge for Use' (K4U), uses philosophical approaches and social science case studies to develop theory and evidence to fortify social policies. In social science, the dynamics examined here tend to be analysed without the label of epistemic injustice. For example, Kaba and Sooriakumaran (2007) describe the shifting power dynamics in the doctor-patient relationship; and Godwin (2011) questions how well-placed the trust in clinical (specifically psychiatric) opinion actually is in judicial and public policy decisions. We present an empirical exploration of how health and social care integration creates an arena in which new forms of epistemic injustice are generated. This is done in the spirit of Fricker's original idea and its salience for decision makers at the coalface of modern health and welfare systems.

Origins of Working Well's Social Epistemology
The two key architects of the UK's most radical recent mental health and welfare reforms, Lord Layard and Lord Freud, both reference the intersection of work and wellbeing, albeit with a difference of emphasis. In Layard's original proposals, welfare-to-work was a key component: For most people with mental problems, activity is an important path to recovery, and work (where it can be managed) is one of the most therapeutic activities. (Layard 2005: 23).
Conversely, Freud quotes the economist and social reformer, William Beveridge, in elaborating what he had always intended should become a requirement for receiving support from the state: "Most men who have once gained the habit of work would rather workin ways to which they are usedthan be idle … But getting work … may involve a change of habits, doing something that is unfamiliar or leaving one's friends or making a painful effort of some other kind…" (quoted in Freud 2007: 2).
However, 'welfare-to-work' is vulnerable to ideological, commercial or political subversion. The Department for Work and Pensions' (DWP) 'workfare' scheme was criticised for asking jobseekers to work for free in a high street retailer (Garland 2019: 102). Nevertheless, the link between work and wellbeing remains a powerful motivator for initiatives that aim to integrate welfare and mental health services. For Layard and Clark (2014: 13), "where the welfare state has gone wrong [is it] had assumed all the problems are external" and it has neglected to address the extent of depression and anxiety that underlies a vast amount of social misery, including unemployment (Weich and Lewis 1998;WHO and the Calouste Gulbenkian Foundation 2014;McManus et al. 2016).
Layard and Freud's analyses led to different perspectives on welfare reform: Layard preferred a CBT-first model of intervention delivered via a national IAPT (improving access to psychological therapies) programme: If our aim is to empower people to control their own lives, psychological treatment has to be an important option… If unemployment was once the most prominent source of misery, it has been replaced by mental illness. (Layard 2005: 3 & 9). While Freud championed a Work-First model with incentives for employment support providers to prioritise skills gaps in their return-to-sustained-employment programmes: A key challenge is to bridge the gap between the "work-first" strategies which have been found to be effective and the shortfall in skills that is evident in the UK economy. (Freud 2007: 53).
Under the Coalition Government, it was recognised that pursuing both policies in parallel, rather than in a more integrated way, was not working. The Ministers responsible (Lord Freud at the DWP and Norman Lamb MP at the Department of Health) jointly commissioned RAND Europe, an independent research institute that aims to improve policy, to revisit the evidence and propose some solutions. 2 RAND identified that: …services were disconnected, often worked in isolation and towards different objectives, and they 'did not speak the same language'. (Van Stolk et al. 2014: 25).
2 One might query the role of an external corporate consultancy in determining such a significant policy decision. The concerns outlined by RAND align with broader integration agendas so are unlikely to have been the sole determinant underpinning the programme. The RAND study happened prior to the inception of Working Well, so remains something of a 'black box'. It would be interesting to gain some insight into how 'paid for' knowledge is valued and weighed against prior assumptions, and especially in how this influenced the programme designbut that is beyond the reach of our inquiry.
IAPT's CBT therapists had some limited success in moving people with mental health conditions off benefits, while the Work Programme's job coaches supported a small number (<0.16%) of claimants with diagnosed mental health problems into work (ibid.: 19-23). However, despite their overarching remits to improve wellbeing and employment outcomes, both services took a narrow focus and essentially failed to acknowledge each other's role.
The Working Well pilot was designed to provide an integrated, more personalised model that addressed the vicious cycle of unemployment and mental health issues. The programme centred on employment support keyworkers, who would connect clients with a range of services, including quick access to talking therapies. Keyworkers and talking therapists would work together to support clients into long-term (>12 months) employment. The shared task was to address barriers to employment for people with common mental health issues, mainly depression and anxiety. Effective team working would enable clients to become better equipped to find work, for example, by having more confidence at job interviews or increased resilience. Keyworkers could be attached to their clients for up to two years, calling on talking therapists as appropriate, with a further 12 months of support for those who entered employment so they could succeed in their new role. The epistemic virtues of Greater Manchester's Working Well pilot therefore derive from close attention to evidence, local knowledge, and experience of implementation flaws in existing government policy.

Methods and Ethics
Our evaluation of Working Well aimed to uncover the causal relationships that led to effective joint working, in particular the moderator and mediator variables in both successful and unsuccessful outcomes. Ethical approval was granted by Durham University on 18th April 2017 and by the London School of Economics and Political Science on 9th August 2017. The study protocol was agreed with Programme Managers at Greater Manchester Combined Authority (GMCA) on 28th September 2017. Data collection ran from October 2017 to May 2018, with a final evaluation report submitted to GMCA in November 2018.
Using a Grounded Theory approach, 14 facilitated discussion, focus group and verification group meetings were carried out with 13 keyworkers and 10 talking therapists, plus three subsequent meetings with managers. This approach allowed us to build a rich picture of the casework and for several themes to emerge, ranging from technical issues such as poor software or inadequate policies around information sharing, to more nuanced issues around different agents' perceptions of each other's work, and how they reflected on their own work. The meetings and interviews were recorded, anonymized, and transcribed. Transcripts were inputted into NVivo (qualitative data analysis software) and coded using Braun and Clarke's (2006) method of thematic analysis.

Practitioners on the Front Line
To contextualise the epistemic dynamics discussed here, it is useful to understand the background for the three agents involved on the front line: Keyworkers (KWs) were the case leads. They were employed through both council-run and private employment services providers, and were the main contact for the client. KWs acted variously as a confidante, advocate and/or sign-poster to other services. KWs saw clients weekly, often befriending them as well as identifying their support needs and arranging the appropriate help for them to become 'employment ready'. They were generally experienced in a range of job types, with lots of knowledge around employment and welfare, as well as 'street knowledge' about local services and practical solutions to everyday issues. An important aspect of the keyworker role was their relative autonomy to select and recommend various options available to clients, one of which was talking therapies. Essentially, they were 'wise', pragmatic, and operated as a kind of 'hub' or 'secure base'.
Talking therapists (TTs) were Psychological Wellbeing Practitioners or High Intensity CBT therapists, employed by the NHS. Clients were referred to TTs for quick access to CBT, ostensibly to address specific mental health barriers to employment. TTs were generally younger and more highly educated than KWs (most had degrees and specific MH qualifications). They did not lead on cases but were considered a valuable collaborative resource, working in tandem with KWs and communicating relevant information about the client's mental health with a shared aim of preparing them for employment. In theory, the TTs were designed to function much as an IAPT team, with the same range of skills in delivering NICEapproved therapies. In practice, they were more effective when they found ways to work flexibly in these roles, often with complex non-employment related issues.
Clients generally had a long history of unemployment and it became clear that some had entrenched negative perceptions of the benefits system. Trust in new initiatives was low and some clients were selective around which information they shared with whom. Much of this was thought by KWs and TTs to be motivated by a fear of benefits sanctions, which fuelled a general distrust in the system as a whole.

Epistemic Injustice on the Front Line
It is worth prefacing our specific examples with some qualifying remarks: First, all focus group participants appeared committed to reflecting on these issues and confronting their disagreements in a spirit of honesty, openness and a willingness to resolve misunderstandings, some of which were addressed directly through changed work practices. Second, although the problem of clinical knowledge being unfairly privileged over other forms of knowledge exists elsewhere, TTs' knowledge was also highly valued by clients and KWs alike. It is not our intention to disparage clinical knowledge but instead to recognise its value within a professional community that accounts for all forms of knowledge.

1) Keyworkers' misplaced humility
KWs routinely deferred to TTs' decisions. This was rationalised through numerous KW comments along the lines of 'they're the experts…' or 'we don't know anything about mental health'. But KWs were experts in their clients' wellbeing and were best placed, by design, to make joint decisions with clients about which services they accessed. KWs' focus on formal clinical training they themselves lacked was overemphasised and detracted from their status as case leads, with TTs serving an ancillary role.
For example, on the tangible issue of making a client referral to the Talking Therapies team, only to have the referral returned as "not appropriate", often without any reasons given, KWs felt disinclined to challenge TTs even when it was clear that their client had been set back by this decision. The injustice was compounded and reinforced by managers, who felt that the solution was for TTs to train KWs in making 'better' referrals. Within the safety of a KW-only focus group, it became clearer why this humility was misplaced and, for some KWs, a resentment lay behind their apparent humility. First, KWs began to question their clients more closely if they had spoken with TTs by telephone or at assessment and been rejected. It emerged that some clients were more willing to open up about mental health issues to KWs and wanted to talk to their KW about these issues rather than to some as-yet unknown individual. Second, where KWs had made a considered judgment that their client would benefit from talking therapy, based on better knowledge of the client's circumstances and a view that they were ready to be challenged within the context of some focused therapeutic work, KWs were properly fulfilling their remit to their clients to advocate and access that support on their behalf.
This 'self-directed' identity prejudice might originate from a pre-existing bias held by KWs: 1) Formal qualifications and clinical knowledge are commonly prioritised in other healthrelated contexts, for example, in the relationships between doctors and nurses, nurses and healthcare assistants, or critically, between healthcare assistants and patients, a relationship that is routinely silenced by hospital systems (Morey and Steven 2019). 2) Psychological services have historically been difficult to access in the UK, 3 contributing to a perception that such knowledge has rarity value compared to other public services. 3) This is emphasised by the highly visible apparatus that surrounds talking therapy services, including qualifications, clinical supervision, strict guidelines and policies, and specialist terminology (which perhaps became more overt when KWs joined the programme).
There is also evidence that KWs' self-directed identity prejudice developed as a response to the particular joint-working environment created by Working Well: 1) The workshopstraining KWs in referral criteria and basic psychological assessmentestablished the clinical process as being the 'correct' and only referral approach, and their didactic element skewed the dynamic between TTs and KWs. 2) When we raised the prospect of KWs feeling diminished in their knowledge, TTs showed some understanding but also some defensiveness, with one commenting 'but we are the experts'. Although some debate is to be expected, that attitude betrayed a sense of knowledge dominance even if joint-working was pursued in good faith.
KW attitudes were not observed prior to commencing the programme but our findings in this study indicate their self-directed identity prejudice was based on some pre-existing bias exacerbated and compounded by their interactions with TTs.

2) Smothering and silencing of the 'expert' keyworker
Even the IT systems used in Working Well sometimes 'made decisions' about clients based on automated responses that overrode the detailed knowledge held by KWs. For instance, a client was 'exited' for non-attendance because the system's calendar appointment schedule could not be changed and so it mistakenly assumed the client did not attend. In itself, this is a straightforward technical error but by stifling certain vital information, including qualitative details of the kind KWs would often garner, the software silenced conversations between KWs and their clients. Ironically, given the integrated rationale for the programme, to enable different disciplines to 'speak the same language', the fact that separate databases used by KWs and TTs didn't communicate effectively further contributed to a knowledge imbalance. 4 In many ways the most critical aspects of expertise were held implicitly, often in the working alliance between KWs and clients. This was systematically smothered by reliance instead upon crude symptom scores from clinical tools designed for TTs. For example, the standard clinical referral tools (PHQ-7/GAD-9) do not capture the detailed social circumstances clients discussed with KWs that might usefully inform TTs about complex problems or suitable joined-up approaches.
Similarly, many clients struggled with drug and alcohol issues, as well as with depression and anxiety. TTs were uncomfortable accepting such clients into therapy without reference to their own referral and assessment criteria. The overall effect was of a clinical 'veil' being drawn over KW/client relationships and all its rich potential insights. While this gave TTs a form of information-based dominance in their three-way partnership, it deprived them of access to more intimate forms of knowledge. Such instances spoke of epistemic injustices as wasted opportunities for greater engagement across more clients, more creative ways of joint working, and the loss of what might have been shared experiences of learning to support each other in helping clients. Compounding the issue, KWs were disallowed from sharing their concerns or expanding on details with TTs due to the strict NHS-inherited confidentiality policies in place to safeguard clients.

Epistemic Injustice among Managers and Commissioners
We also interviewed managers and commissioners, in the expectation of obtaining additional perspectives and some explanations for why aspects of the programme functioned as they did (or not) from the people directly responsible for making those decisions. Most of the managers had been involved with Working Well from the start and the commissioners had been part of the original design team. Yet it was among these respective leadership roles that we encountered something unexpected, namely, a level of epistemic violence (Dotson 2011) that appeared to have soured working relationships in the regular performance meetings to the extent that communication had clearly broken down. Again, it helps to contextualise what we identify as epistemic violence by explaining the positions and roles of the main agents.
Managers were responsible for delivering the Working Well service. Two provider organisations employed the KW teams; one local authority owned and run, and the other a commercial employment support provider. These managers had various professional backgrounds, including delivery of NHS-funded services and integrated care programmes, and provided direct supervision for KWs and team leads. The talking therapies team was employed by a large mental health NHS Foundation Trust and their managers were themselves senior talking therapists with clinical backgrounds, directly involved in clinical supervision as well as performance management.
Commissioners had lead strategic roles for employment and skills across GMCA. They worked closely with public health and mental health commissioners, and reported to a Local Authority Chief Executive. As well as their formal role in steering performance, the commissioners had a hands-on role in facilitating the way that Working Well connected to the wider system of health and social care services. In many ways, the "integrated" aspect of the programme was dependent on a wide set of relationships and networks these commissioners facilitated.
We do not think we are overstating the breakdown in communication between managers and commissioners by using the term 'epistemic violence', defined as… …a failure of an audience to communicatively reciprocate, either intentionally or unintentionally, in linguistic exchanges owing to pernicious ignorance. (Dotson 2011: 250).
There were two lines of communication difficulty, which all interviewees acknowledged, though each from their own perspectives. First, NHS managers and the commercial provider managers spoke a 'different language', resulting in what appeared as 'identity prejudice'. Animosity on the part of NHS managers may be a reasonable response to the sense of creeping privatisation that has been a concern over recent years. Animosity on the part of the commercial provider's managers is likely to be linked with frustration at the perceived inertia and inefficiencies of public services. However, we were surprised by the vehemence with which the opposing views were held. Both the commissioners and the managers from the council-run provider found themselves in the middle of continual disagreement between the NHS and commercial managers.
Second, notwithstanding the purpose of Working Well's integrated design to pre-empt silo working and to foster a sense of shared endeavour, it was clear that the NHS managers were overly focused on clinical outcomes rather than on employment and skills outcomes. The commissioners noted that it might have been better if we had been present during their performance meetings, so that the kind of reflection that had been observed among KWs and TTs might have been possible at manager and commissioner level.
One specific examplethe commercial provider investing in their own in-house counsellor part-way through the delivery of the programmecaptures the essence of these noncollaborative socio-epistemic practices of mutual 'silencing'. We set out the perspectives separately, hopefully doing justice to each.
1. The commercial provider's justification for commissioning their own in-house counsellor was: a) some GP referrers were requesting counselling for their clients (e.g. where bereavement was part of the picture of ongoing depression and barriers to employment); b) the Talking Therapies team only offered CBT despite the 'menu' of NICErecommended options that was originally intended to be available 5 ; c) it was increasingly clear that the NHS managers had no intention to enable this provision; and d) providing this offer in-house was shown to reduce waiting times and improve referral take-up, and was expected to improve eventual employment outcomes. 2. NHS managers disputed the need for a counselling option, claiming that a) their own team could offer the appropriate help for clients, as demonstrated by the excellent recovery rates they were achieving; b) the best predictor for success was not the type of therapy on offer but the aptitude or match between client and TT, whose skills were not in question; and c) the counsellor employed by the private provider was not properly trained to deliver evidence-based therapy (NHS TTs had attended some of that counsellor's workshops). This was considered to be offering counselling 'on the cheap', thereby short-changing Working Well clients. One NHS manager commented: 'you have to know what you are dealing with and what you are doing… this is a dis-enfranchised group; we believe they deserve the best' (defined by NICE as evidence-based CBT). 3. Commissioners and managers from the council-run provider could see both sides in this argument. They defended the commercial provider against the accusation from the NHS managers that 'you are trying to prove you can deliver Working Well on your own without the need for evidence-based mental health expertise because you want to win more contracts and make more profits for your shareholders'. Personalisation was a core principle of Working Well, so offering more options was deemed a legitimate way of enhancing the potential benefits of the programme. Doing this ahead of scaling up Working Well, even if this also gained the provider a competitive advantage, was considered reasonable. Equally, the commissioners were perfectly happy that the NHS managers had recruited a team of CBT-based TTs, given that NICE guidance recommended CBT across a wide range of common mental health disorders. Commissioners knew that the NHS managers had been unable to source any multi-skilled candidates locally but it was feasible for the TTs to develop additional skills as the team evolved. Further, the TT team were providing a high quality, reliable, safe, effective and flexible service.
Investing their own money in a resource that enabled them to provide a more effective service improved the commercial provider's chances of being recommissioned. Claiming there was a GP demand for clients to access more diverse forms of counselling helped legitimise this strategy but also raises the possibility that gaps in the programme could have been misrepresented for commercial gain. That said, in our study, TTs spoke of some clients struggling with bereavement and other complex social problems that may have benefited from less directive approaches, such as counselling. The question of whether the GP demand was significant is important but remains unanswered. Whether genuine or not, deploying protected (client) data and appeals to authority (GPs) betrays a strategic use of the epistemic terrain. Moreover, the claim implies that the programme itself fails to provide adequate hermeneutic tools to help all clients make sense of their issues. By proposing to broaden this offer and taking on the investment risk, these actions raised difficult questions around the joint commissioning of public and private organisations. This position, employing an in-house counsellor, could not be adequately reconciled into the shared plan already agreed with the NHS and council-run providers, and the problem of public and private organisations 'speaking different languages' remained at the mid-and higher levels. The NHS has been described as a 'socialist island in a capitalist sea' (Powell 1997: 187) an important ideological discrepancy that gave rise to mutual silencing and wilful resistance to knowing on both sides of the debate. The competition between NHS and private providers, not just to be recommissioned here but also to establish power in a developing integrated landscape, caused rifts that underlined a vicious element between providers that went beyond ideological differences. In programme terms, it poisoned the joint working relationship at management levels, as illustrated by some of the NHS managers' comments. In commissioning terms, it detracted from the integration focus. These are more intractable problems compared with the pragmatic solutions found by the frontline workers, highlighting the contrasting inflections of epistemic injustice that emerge at different points in a system. When the contracts were decided for the next phase of scaling up, the two KW provider organisations won the bid and the talking therapies team was disbanded and redeployed elsewhere. The Working Well Board decided that it did not need dedicated mental health input. In this instance, the clinicians lost out, contrasting with the position at the front line and revealing a tension that both privileges and de-privileges the clinical position in relation to other professional domains at different layers in the system. We were not given access to Board-level decisions but attended the Board meeting that discussed our report and were informed of their decision at that meeting. The reason stated was that funding for the pilot stage had been a one-off combination of Cabinet Office and European Social Fund monies. That said, in scaling up and continuing with Working Well beyond the pilot, decisions had been taken locally as to how and what would be funded. As we were not party to those discussions our comments here must be considered speculative, although there is no obvious connection between the Board's decision and our evaluation report. It was noteworthy however, that the Mental Health Commissioner was not present at the Board meeting, which we took to be symptomatic of a failure, not of the pilot or its providers, but of Greater Manchester's commitment to joined-up commissioning at a strategic level. In concrete terms, local budgets for employment and skills, and for mental health remained separate. The existence of separate budgets with separate commissioners does not preclude joint commissioning and, indeed, learning from the pilot may well involve this in a later iteration of Working Well, as the commissioners themselves expressed to us. However, without reform at a more strategic level, silos will still operate in commissioning decisions.

Species of Emergent Epistemic Injustice Observed in Working Well
In her call for the construction of virtuous social systems, Anderson sees integration as a vital structural organising principle: If group segregation is the structural ground of … epistemic injustice … then group integration is a structural remedy-a virtue of epistemic institutions … Shared inquiry also tends to produce a shared reality, which can help overcome hermeneutical injustice and its attendant testimonial injustices. (Anderson 2012: 171).
Our elaboration of emergent species of epistemic injustice, generated by structural features of a system or institution, appears both to confirm Anderson's assertion at one level, while disconfirming it at another. Among frontline practitioners there were clear disparities of status in terms of knowledge and value given to expertise. This happens in other professional contexts but in Working Well, all practitioners faced the weight and depth of their clients' complex problems (including in some cases, problems caused by the welfare system itself). The disparities we observed increased the risk of programme drift, towards solving mental health issues rather than taking a collaborative approach to achieve employment outcomes. One could see this as reflecting an organic identification of priorities, not least since the employment intervention was framed as being partly oriented towards assisting with mental health issues. But the underpinning mid-range theory was to break the cycle of mental health issues and unemployment. While the programme viewed both issues as being inextricably linked, its tangible aim, expressed as a target, was to help its clients into good quality long-term employment and keyworkers, rather than clinicians, were the first port of call.
It could be argued that by prioritising an employment target as an outcome (and which the costbenefit analysis for Working Well depended upon), this simply replaces one epistemic hierarchy with another. Likewise, to paraphrase Goodhart's Lawwhen a measure becomes a target it ceases to be a good measureit could also be the case that epistemic injustice operated at other levels of the programme, and in other ways, than those we were able to observe. What we can say is that a pathocentric epistemic complex emerged at this front line level in spite of the design features that were intended to mitigate against this: namely, that the stated offer to clients emphasised voluntary participation with the aim of employment as an outcome, and the shared target across the providers also gave them equal responsibility for delivering employment outcomes.
It should also be recognised that, alongside its targets at programme level, a personalised approach was designed for each client to track subjective wellbeing and personalised clientcentred goals. These were operationalised through use of an idiographic tool that allowed each client to define their own recovery goals in their own words. Given the heterogeneity of clients' needs and circumstances, Working Well's design allowed flexibility in prioritising employment and/ or MH outcomes as appropriate, with space for KWs, TTs and clients to discuss and agree which specific needs should be met first, en route to achieving that client's goal.
We have indicated several structural and organisational issues that tended to reinforce the knowledge asymmetries (Carel and Kidd 2017: 336) between KWs and TTs, and entrench the privileging of clinical expertise over other knowledges. These were further crystallised by programme-specific elements, which filtered out the more nuanced experiences of both KWs and clients, supressing non-clinical voices and instituting a language for talking about mental illness that gave a hermeneutical advantage to the clinicians. But is this so bad? TTs were employed after all as highly-qualified specialists. Conversely, KWs were employed on the basis of their experience; some having been welfare recipients, some having experienced their own mental health difficulties, and all having 'street knowledge' of local challenges and resources. Both brought valuable epistemic resources into the programme. That TTs and KWs could be categorised by these unequal forms of knowledge, inevitably they would feel valued differently, and this would naturally be reflected in their routine inter-disciplinary team dynamics.
While the value of accessing different types of expertise in integrated care might vary for clients at different times, those in advocacy roles occupy a critical area for promoting epistemic and consequently social justice by securing their client's entitlement (Newbigging and Ridley 2018). By habitually deprecating their lack of 'expert' knowledge, KWs diminished their intended advocacy role. Instead, they deferred to the TT's 'assessment', when their actual request on behalf of their client, appropriately, had been to start a course of one-to-one therapy, not for an assessment. We do not wish to claim a special status for these solely based around identity prejudice, but by drawing attention to what is being protected, namely, defensive ways of working, we hope to shed new light on addressing this.
A culture of protecting professional and organisational identities is one of the most prominent barriers to new ways of working, especially where established skills and roles are reconfigured. Other barriers include overestimating the capacity of individual roles to deliver integrated care, difficulties in making these roles sustainable over time, and poor accountability and oversight of staff in roles that do not fit into established structures. (Gilburt 2016: 3).
Due to the sheer inertia of the doxa depression-as-disability among the client group, that was way beyond the resources of any individual practitioner or client to be able to shift, we believe KWs and TTs began to see increasingly that they needed to integrate better collectively. Arguably, participation in our evaluation, as a shared inquiry on equal terms, generated what Anderson (2012: 171) refers to as a 'shared reality', which helped both the KWs and TTs to rebalance their team dynamic. In Aristotelian terms, there began to be a greater appreciation for phronesis, or 'practical wisdom', as practised skilfully by KWs in engaging and supporting their clients. Likewise, a more realistic sense of the limitations of formal learning, in a context where considerable flexibility and adaptability was needed, began to emerge. This might be characterised as integrating episteme with phronesis. In some cases, pairs of KWs and TTs simply invented for themselves new ways of joint working with clients. More generally it was clearer to them all that certain moments of emotional availability with clients could fall to each of them, and that sometimes this required extra time and space to support. In other words, good inter-disciplinary relationships evolved to replace protocol-led ways of working.
At the level of managers and commissioners there was a different story: epistemic violence got worse, until one group was eliminated. We were not present during performance meetings with managers and commissioners so we cannot really know why this forum for shared inquiry led towards a vicious rather than a virtuous outcome. However, we speculate that a combination of different ideological positions and an inherent competition between the NHS and the commercial provider played a role in the animosity that emerged between them.
There was also a vicious structural element in Working Well at the welfare policy level (Battaly 2013(Battaly , 2017. A vicious policy may be a structural organising feature of an integrated programme that promotes vicious actions, attitudes or states of affairs among its participants (Kidd and Carel 2019). Time and again we heard about practitioners' efforts being undermined by benefits sanctions, and a fragile sense of trust in the whole system once again being shattered. While all participants were aware of the impacts of austerity on Working Well clients, it was psychological therapists, and their clinician NHS managers as regulated professionals, who were most acutely aware of the unethical (vicious) implications of benefits sanctions (see Consensus Statement on Welfare Sanctions 2019). If Working Well meant what it said, then the work involved for all participants on the programme (clients, practitioners, managers, commissioners) needed to be 'good work'. Good work is not to be found at the sharp end of the threat of benefit sanctions.
In January 2020, a parliamentary question was raised regarding links between mental health and benefits sanctions. The DWP minister responded that "no assessment had been made" of such links and mentioned no future plans to do so (Parliament.uk 2020). This is a prime example, directly relevant to the concerns raised in the consensus statement, of the highest levels of government silencing the clinical community.

Conclusion: Challenging Pathocentric Epistemic Complexes
Integrated care, where different professions must work together, provides an arena in which multiple forms of epistemic injustice play out. Many of these are caused by systems rather than by individuals and our evaluation included some suggestions for challenging these. First, recognising the different but equally valid forms of knowledge held by both KWs and TTs must form the basis of the joint working relationship. Happily, our joint reflective meetings enabled the different disciplines to work together on equal terms, so we asked for these to become a regularised event. Second, managers and commissioners need to be able to acknowledge each other's different positions and find ways to accommodate these in a spirit of mutual understanding, honesty and respect. Third, such initiativeswhich are designed to help peoplemust operate within a policy framework built on similarly virtuous principles. This requires, in the first instance, government acknowledgement of the evidence placed before it.
In Working Well, a pathocentric epistemic complex, governed by arbitrary ideological and/or competitive influences and underpinned by a vicious policy, converged to produce new forms of epistemic injustice. This generated a tension in which clinical knowledge at the front line was unfairly but unintentionally prioritised, yet NHS services lost out to local authority and private providers at the commissioning level. While this contradiction might not be replicated in all integrated contexts, the complexity of such systems makes it possible, damaging outcomes for clients (adding to their existing problems) and reducing confidence in joint working. Given the combination of injustices that arise at the intersection of mental health and welfare services, and the increasing integration agenda in the UK and globally, this seems like a perfect storm. Yet, integration remains one of the most promising avenues for public services. Drawing on learning from this study, we therefore caution against creating environments in which damaging epistemic factors can work against the overall goals of integrated programmes.