Diamond in the rough? An “empty review” of research into “neurodiversity” and a road map for developing the inclusion agenda

Purpose – The aims of the paper were to highlight the dearth of applied practitioner research concerning the expression of neurodiversity at work and develop an epistemological framework for a future research agenda. Design/methodology/approach – A systematic empty review protocol was employed, with three a priori research questions, inquiring as to the extent of neurodiversity research within mainstreamwork psychology, psychology in general and lastly within cross-disciplinary academic research. The results of the final search were quality checked and categorized to illustrate where studies relevant to practice are currently located. Findings – The academic literature was found to be lacking in contextualized, practical advice for employers or employees. The location and foci of extracted studies highlighted a growing science-practitioner gap. Research limitations/implications – The research focused on common neurominority conditions such as autism and dyslexia; it is acknowledged that the neurodiversity definition itself is broader and more anthropological in nature. A need for a comprehensive research agenda is articulated, and research questions and frameworks are proposed. Practical implications – Guidance is given on applying disability accommodation to both individual and organizational targets. Social implications – The disability employment gap is unchanged since legislation was introduced. The neurodiversity concept is no longer new, and it is time for multi-disciplinary collaborations across science and practice to address the questions raised in this paper. Originality/value – This paper offers an original analysis of the neurodiversity paradox, combining systematic inquiry with a narrative synthesis of the extant literature. The conceptual clarification offers clear directions for researchers and practitioners.


Introduction
"Neurodiversity" broadly refers to naturally occurring diversity in human cognition (Singer, 1999). It has been used as an umbrella term for a range of neurocognitive developmental disorders (Doyle, 2020;Kapp et al., 2013) including autistic spectrum disorder, attention deficit and hyperactivity disorder or dyslexia. Recent public dialogue has shifted to the term "neurominority" to signpost that relevant individuals are disadvantaged regarding a range of life outcomes, including systemic social exclusion in education and inferior employment outcomes (Carter et al., 2012;Snowling et al., 2000). Subsequent restriction of opportunities for a fulfilled working life and An empty review of neurodiversity research career (Holliday et al., 1999;Taylor and Walter, 2003) is in contrast to emerging popular narratives about the talent potential of neurodiversity for modern workplaces (Austin and Pisano, 2017;Sniderman, 2014): the "diamond in the rough" . In this paper, we highlight the inadequate scope and focus of academic attention on neurodiversity and employment to date across the fields of applied psychology and management studies. We contextualize the neurodiversity paradigm from a medical, social and legal perspective in the wider workplace equality, diversity and inclusion agenda. Through a series of "Empty Reviews" defined as a targeted analysis of literature gaps through systematic literature review principles (Schlosser and Sigafoos, 2009;Yaffe et al., 2012), we illustrate the dearth of tangible evidence. Through due diligence in our conceptualization, including definition of gaps in knowledge, we outline a future research and practice agenda to facilitate informed inclusion for a marginalized minority. In other words, it is our aim to support evidence-based practice in this emerging field guiding human resources (HR) and employment policy to mitigate labor force exclusion.

Defining neurodiversity
The term "Neurodiversity" was coined by the sociological researcher Judy Singer (1999) in the context of rights activism during the 1990s (Mcgee, 2012;Runswick-Cole, 2014). The neurodiversity movement highlights the life-long and positive aspects of naturally occurring cognitive "differences", such as creativity and "special interest" skills (Meilleur et al., 2015;Von K arolyi et al., 2003;White and Shah, 2006), as opposed to the focus on developmental "deficits" such as language or processing speed (Armstrong, 2010;Grant, 2009;Jurecic, 2007;Kapp et al., 2013). Neurodiversity has been considered a progression from previous umbrella terms such as specific learning difficulties, neurodevelopmental disorders or hidden/invisible impairments. However, Ms Singer herself, alongside other researchers and advocates, prefers "neurodiversity" to refer to the cognitive diversity in all humans (Chapman, 2020;Monzee et al., 2019;Singer, 1998;Walker, 2012) rather than appropriated as a synonym for disability. Relevant terminology continues to evolve and be hotly debated; we briefly define alternatives that denote equivalence to the outdated deficit model language. Those who fall in a statistical norm, based on relevant cognitive tests or behavioral assessments, are referred to as "neurotypical," and those who would previously have been termed disordered are referred to as "neurodivergent," "neurodiverse" or part of a "neurominority" (Bottema-Beutel et al., 2020;Doyle, 2020;Singer, 1998;Walker, 2012). Some authors include general learning disabilities and/or mental health needs in definitions (Armstrong, 2010;Chapman and Veit, 2020). Typically, though, the following conditions are considered neurominorities: attention deficit hyperactivity disorder (ADHD), autism, dyslexia, dyspraxia/developmental coordination disorder (DCD), Tourette syndrome (TS), dyscalculia and dysgraphia (Bewley and George, 2016;Grant, 2009;Jurecic, 2007;Kapp et al., 2013;Snowling, 2005). Definitions of each condition have evolved over time, and diagnostic criteria, as well as treatment approaches, continue to vary considerably (Elliot and Grigorenko, 2014;Kirby et al., 2011;Shelley-Tremblay and Rosen, 1996). For a historical review of neurodiversity/neurodivergence since the industrialisation, see Doyle (2020). Below we present a brief summary of some of the educo-medical models and conceptualizations and prevalence; see Doyle (2017) for a working summary of the occupational strengths and weaknesses of each condition.
The medical model of neurodiversity. This model defines neurominorities linguistically through pseudo-medical deficits. The Diagnostics and Statistical Manual of Psychiatric Disorders (American Psychiatric Association, 2013), known as the "DSM-5", is used by relevant professionals to categorize and determine individual cases. These conditions are developmental as they emerge in childhood and/or adolescence, but not in response to trauma or ill health and confer specific difficulties rather than indicate global developmental delay (Snowling, 2005). Psychological diagnostic criteria define the hallmark of a neurominority as EDI a statistically significant within-person difference, showing a disparity between cognitive strengths and weaknesses (e.g. verbal/visual reasoning and memory). This is in contrast to a "neurotypical" presentation where an individual's ability scores are within a standard deviation or two of each other (either below average, average or above average) (Grant, 2009;Ihori and Olvera, 2015). There are two sub-types of developmental neurominorities (Doyle, 2017). Population prevalence is shown in parentheses, based on best available data, noting that these vary across nations and depend on which diagnostic criteria are applied: (1) Clinical conditions, which are assessed by behavior/communication and typically diagnosed by psychology, neuropsychology and psychiatry clinicians working in health services: ADHD around 5% worldwide (Catal a-L opez et al., 2017;Shelley-Tremblay and Rosen, 1996) but higher in USA (Danielson et al., 2018).
Autism between <1 and 1.6% globally, significantly affected by diagnostic criteria and access to services (Elsabbagh et al., 2012).
TS 1% (CDC, 2009;Robertson, 2006) (2) Applied conditions, which are assessed by functional difficulties, either educational or occupational, and typically diagnosed by applied psychologists, educators and occupational therapists in school or work settings: Dyslexia up to 10% (Snowling, 2010).
For the purpose of this paper, we focus on the main developmental neurominorities, namely ADHD, autism, dyslexia, DCD and Tourette syndrome, to gauge the quality of research regarding practical support and inclusion of these conditions in the workplace. The emerging social model of neurodiversity. Given the high prevalence of neurominorities reported in advanced economies, there is an obvious evolutionary critique that neurodiversity has evolved within a typical spectrum of human experience (Blank et al., 2008;Boycott et al., 2014;Doyle, 2020;Shelley-Tremblay and Rosen, 1996). As it may be natural and useful to have a small percentage of the population with specialist rather than generalist abilities (Armstrong, 2010), stakeholders argue that we need to develop neutral inclusion practices that do not insinuate ill health. It is worth noting that during the past 100 years both left-handedness and homosexuality were considered psychiatric disorders, demonstrating that any conception of "normal" is culturally bound and subject to historical changes in line with ecological systems theory (Bronfenbrenner, 1979). We cannot be sure at this stage that a disorder based on quantifying the degree of adherence to a social norm, such as literacy, activity levels or social interactions, is truly a biological deficit. The social model of neurodiversity contends that the world is polarized and lacks in flexibility to accommodate all natural variations in human cognition and functioning (Shakespeare and Watson, 1997). Individuals who fall outside the norm represent a minority group, whose rights must be protected and whose ways of functioning should be recognized, valued and harnessed (Runswick-Cole, 2014;Singer, 1998).
The social model has influenced public awareness of neurodiversity in recent years; the narrative of the "diamond in the rough" is voiced in popular press and media worldwide (Adams, 2019;Philipson, 2014;Sniderman, 2014;Wollaston, 2016). Neurodiversity is also attracting the attention of the business press who describe a "talent advantage" for employers through targeted recruitment of neurodiverse applicants (Austin and Pisano, 2017;Comaford, 2017). This approach chimes with broader initiatives to promote diversity and inclusion of gender, ethnic minorities and sexual orientation (Cucina et al., 2013) and disability more generally (Murfitt et al., 2018) to broaden the talent pool, known as the "business case" for inclusion (Saleh and Bruy ere, 2018).
Socio-legal barriers and protections. There is a conflict between aspiration and reality in occupational practice. Despite sparse yet consistent evidence in support of specific talents (Armstrong, 2015;Logan, 2009;Meilleur et al., 2015;Von K arolyi et al., 2003;White and Shah, 2006), neurominorities are currently not well accommodated by contemporary social structures based on neurotypical profiles. To illustrate, in the UK, modern apprenticeships require (high) literacy levels of grade 10 or above to access basic vocational training in fields that rely on specialist visual reasoning, such as hairdressing and plumbing (SFA, 2016). Interviews remain popular as a hiring technique globally, despite evidence that these unfairly disadvantage neurominorities, particularly autistic people (Cooper et al., 2018;Hayes et al., 2015). Such barriers matter. Only 10-16% of autistic people have a job (NAS, 2016), and ADHD has been shown to significantly increase the likelihood of incarceration in the UK and USA (Halmøy et al., 2009;Young et al., 2018). Over 20% of UK and 35% of US entrepreneurs are dyslexic compared to only 1% of corporate managers (Logan, 2009), and a third of longterm unemployed people are dyslexic (Jensen et al., 2000). Such systemic exclusion has moral, social and economic consequences.
To counter extant structural barriers in education and work, neurominorities are considered recognized disabilities in most developed nations, which theoretically afford protection from discrimination by law. In the UK Equality Act (Equality Act, 2010), for example, reference is made to long-term (chronic) difficulties in memory, learning and communication which may be affected by all the above-named conditions. In the USA, Canada, Australia and throughout Europe, similar legislation provides disability protections (DESE, 1992;Govt of Canada, 1995;Americans With Disabilities Act, 2008;IDEA, 2006) all broadly aligned to the United Nations Convention on the Rights of Persons with Disabilities (UNCRPD) (United Nations, 2006). Such legislative innovations are yet to translate into occupational success. The "disability-employment gap" remains of concern worldwide (WHO, 2011). There is a pressing need to understand "what works" and move to an evidencebased practice model for applied practitioners to implement, yet this "push" for interventions based on diagnosis somewhat undermines the "pull" of the untapped talent model. Although applied research is sorely needed, ontological positions to date have been dominated by the medical model. Yet individuals with lived experience advocate a more critical and realist approach to inclusion, which takes account of their priorities through inclusive research design (Bottema-Beutel et al., 2020;Rios et al., 2016;Walker, 2012). We return to this point in our discussion.
A science-practitioner gap. Activities that constitute disability "support" vary geographically, with advanced economies applying a range of active labor market interventions supporting both the individual (pre-and during employment, "supply side") as well as the employer through incentivization and best practice guidance ("demand side") (Murfitt et al., 2018;Saleh and Bruy ere, 2018). However, the extent to which such programs are effective for neurominorities is less documented (Gerber et al., 2012). In the UK, for example, there is a statutory government body facilitating the provision of "reasonable adjustments" or accommodations to thousands of employees with invisible disabilities per year; the third most common disability group after muscular-skeletal and sensory impairment (Gifford, 2011). The program, known as "Access to Work", offers one-to-one coaching and assistive technology to individuals (at approx. $1000 per person) and is broadly well-regarded as evidenced by user survey data (Adams et al., 2018;Melvill et al., 2015). Similar programs such as "Jobs in Jeopardy", part of the Australian Disability Employment EDI Service, are also seen to add good value for the tax-payer, yet do not specifically focus on the experience of neurominorities, with the exception of autistic people (DEEWR, 2013). The provision of such interventions has received minimal attention in research evidenced by high level summaries, yet insufficient high-quality studies. We are unaware of any process evaluations of relevant interventions most frequently recommended for neurominorities by these programs, such as coaching, mentoring and assistive technology (Doyle, 2019a, b;Kulow and Thomas, 2019;Lindstedt and Umb-Carlsson, 2013;Sundar, 2017;Work and Pensions Committee, 2018) that could elucidate psychological mechanisms of change or evidence career advancement outcomes. We are witnessing increasing practitioner advice from business advisor groups (ACAS, 2016;ODEP, 2020;Spargo-Mabbs et al., 2020;TUC, 2011) including coaching, assistive technology, provision of extra time, flexible hours policies and use of acoustic barriers to mitigate background noise distractions. There is a lack of reliable experimental work evaluating the effectiveness of relevant support structures for neurodivergent individuals (Patton, 2019;Sundar, 2017;Whitby, 2017), meaning that it is difficult to demarcate evidence-based practice (Briner and Rousseau, 2011).
Globally, disability programs tend to be led by healthcare provision, vocational rehabilitation and occupational therapy tending toward those with more medicalized needs (Doyle, 2021). It is unclear the extent to which neurominorities such as ADHD, TS, dyslexic and dyspraxic employees might be covered by such programs, though autism has received increased attention (Lawer et al., 2009). For example, Individual Placement Support (IPS) is sometimes provided for autistic people, which is well-regarded and reasonably well-evaluated (Lounds- Taylor et al., 2012;Wehman et al., 2016). The principles and practices of such relevant interventions could provide a rich resource for expansion to accommodate other neurominorities. Yet they are based on a medical model, which implies that the workplace environment is fixed, and the individual is impaired and therefore needs to adapt. This is at odds with the social model ambitions of the neurodiversity movement and the UNCRPD (United Nations, 2006). Further, at an average cost of tens of thousands of dollars per person, which is an investment that an employer or employee alone cannot justify, IPS interventions only break even when reductions in public spending are included (Jacob et al., 2015;Knapp et al., 2013). Employee Assistance Programs are also relevant in this context as a potential source of specific support. However, such programs tend to accommodate mental health specifically; they are potentially oblivious to cognitive strengths and impairment, as well as being somewhat vague about any benefit for individual work performance (Joseph et al., 2018). While it may be essential to address acute mental distress at source, it is possible that support for stress and anxiety when the cause is cognitive (not emotional or social) could create more long-term harm than good if support activities are not informed by a more holistic perspective on neurominorities. More specifically, any presupposition that the cause is individual failure to adapt rather than structural exclusion could be detrimental for self-worth, employment and other outcomes, leading to a spiral of negativity. Therefore, any academic enquiry must take a holistic perspective on interventions to address both socially constructed marginalization and individual needs, in order to inform evidence-based professional practice.
In summary, evidence regarding any activities to support improving job performance and subsequent occupational inclusion through disability accommodation for neurominorities is lacking (Adamou et al., 2013;Kirby et al., 2011;Palmer and Stern, 2015;Rice and Brooks, 2004). We consider the development of a research agenda an essential and urgent task for the neurodiversity inclusion movement.

Review questions
Against a context of overlapping and potentially conflicting societal, legal and conceptual developments, we aimed to map any existing evidence and scrutinize robustness as well as disciplinary origin. Although we endeavored to keep an open mind, it was our assumption An empty review of neurodiversity research from the outset that our research would draw from the principles of "Empty Reviews". Empty reviews are becoming common in health and medical fields to establish lack of evidence for any interventions (Yaffe et al., 2012). Guided by a systematic review methodology, we (1) formulated specific a priori questions, (2) devised a search protocol, (3) applied a predefined relevance inclusion criteria and process for achieving consensus and (4) applied predefined quality inclusion criteria. Our protocol was based on best practice guidance from the management/ business field, given the work context for any relevant evidence (Denyer and Tranfield, 2009;Rousseau et al., 2008). Our research questions were: (1) How many papers are published in mainstream empirical organizational and management psychology publications related to neurodiversity (or related conditions) in work contexts?
(2) How many papers exist in general psychological domains regarding neurodiversity?
(3) How much research exists in any academic domain concerning the occupational inclusion of neurominorities / neurodivergence?
From the third search, we were able to formulate a fourth question: (4) In which academic disciplines is relevant academic research currently located?
From these structured searches, our synthesis will consider: (5) What, if anything, can we infer from the current literature about "what works" in practice to improve the inclusion of neurodivergent individuals at work?
(6) How should the prospective research agenda be epistemologically framed?
Systematic empty review protocol Search terms Our review focused on developmental neurominorities using search terms which reflected differences in labels over time. We excluded mental health needs, chronic health conditions, acquired brain injury and general learning disabilities since they are already addressed in more advanced research fields (Corbi ere et al., 2014;McGonagle et al., 2014;Tyerman, 2012). Some mental health research was extracted as a point of reference to compare volume of returns. Dyscalculia and dysgraphia are included in the taxonomy of developmental conditions, but not considered in the extraction given the previously noted absence of research (Doyle, 2017). Table 1 summarizes all search terms. Extraction protocol. There were three iterative searches conducted in January 2017 and repeated in August 2018; these are summarized in Figure 1.
Search one. How many papers are published in mainstream empirical organizational psychology publications related to neurodiversity (or related conditions) in work contexts?
The first search represented a systematic check that this field was indeed underresearched, by focusing on what ought to be ideal publications for considering the intersection of neurodiversity (a psychological phenomenon) and workplace inclusion: The Journal of Applied Psychology, Personnel Psychology, Industrial and Organizational Psychology and the Journal of Occupational and Organizational Psychology. We commenced by hand-searching the four journal websites specifically for the five conditions and the term "neurodiversity" to understand the extent to which they featured in key applied psychology research.  No time limits were applied. The primary aim of this search was to extract the number of references to neurominorities and the term "neurodiversity" itself, and we did not initially subject papers to a second stage rigorous relevance and quality criteria check. However, the small number of returns enabled a qualitative review of the extent to which conditions were merely "mentioned" as opposed to explored conceptually or evaluated. Table 2 shows the number of returns compared to the number mentioning mild-to-moderate mental health needs in two of the journals. Psychology in order to sample-check whether or not consideration of neurodiversity was better represented in relevant psychological sub-disciplines. The second search was not intended to be definitive, but indicative of the spread of individual condition representation in research relevant to the neurodiversity. We did not apply time limits, but categorized the returns into papers published before 1995, 1995-2005 and 2005þ to consider if the developing social movement was in any way linked to number or foci of publications. As with search one, the extracted papers were not subjected to relevance and quality review and may only "mention" the condition rather than explore thoroughly. We compared the relative percentages of each condition's prevalence rates with the relative volume of research to observe any over-or underrepresentation.

Variations of terms in search
Search three. How many papers address the occupational presentation of neurodiversity to a high academic standard of evidence in any discipline?
Question four. In which academic disciplines is relevant academic research currently located?
We undertook a wider, more international review of the literature on occupational presentation of neurodiversity in EBSCO-hosted journals to ascertain if there was a tranche of yielding principles (Cucina et al., 2013) that could be appropriated for inclusion practice from related disciplines, for example vocational rehabilitation. We limited the search to the English language and included the grey literature using all terms in Table 1. We took a systematic approach, with an initial 1439 returns creating a bespoke extraction form in Excel. We then filtered results by abstract screening, removing any papers without reference to  We did not explicitly seek to apply a rigorous quality check at this stage mainly because the aim of the review was not meta-analysis of an intervention success, simply exploration of the extent of research. Instead, we considered the hierarchy of evidence principles (Rousseau et al., 2008) and used quality guidance from related systematic reviews (Doyle and McDowall, 2019;Rojon et al., 2011) to rate papers as strong evidence (systematic review and metaanalysis), medium evidence (well-constructed primary papers) or weak evidence (poorlyconstructed papers, case studies, practitioner opinion and anecdotal evidence). The primary author performed these ratings, the second author cross-checked and discussion was held to resolve conflicts until full consensus was reached. Papers scoring medium or above were retained: 48 papers in total. Finally, the papers were categorized by academic discipline using journal titles (or full paper review if not clear) to consider where, if not in organizational psychology or management, knowledge regarding the occupational presentation of neurodiversity is currently located. Figure 1 depicts the number of included papers at each stage.

Results and synthesis
Below we present the number of returns in each extraction, categorized as described above, and a brief narrative summary of the findings for each search according to the first three review questions: Search 1: organizational psychology journals Table 2 shows the number of papers mentioning neurominority conditions and an absence of the phrase neurodiversity to date in the contemporary organizational psychology that we selected. Only 32 papers mentioned neurominority conditions at all; 22 of these were in SIOP's Industrial and Organizational Psychology journal, three in the Journal of Applied Psychology and two in Personnel Psychology. The Journal of Occupational and Organizational Psychology (JOOP) mentions neurodiverse conditions on only five occasions compared with 413 mentions of mental health conditions. Full paper review of the returns did not provide evidence of the intervention evaluations required to support disability inclusion practice, protect employers/employees or justify current policy. For example, the two dyslexia returns within JOOP pertain to (1) the use of handwriting as a recruitment technique (Klimoski and Rafaeli, 1983) and (2) a broader article about employer responses to disability inclusion (Jackson et al., 2000). Similarly, the other neurominorities listed above are mentioned in references related to a broader theme rather than examined as to how they relate to occupational psychology per se (for example, both ADHD and autism are within the reference list for the coaching review conducted by Jones et al., 2016). While SIOP's Industrial and Organizational Psychology journal marks a significant improvement in volume compared with others in the discipline, there are two points of note: (1) mental health papers were still disproportionally better represented and (2) papers were cross-sectional and/or conceptual reviews offering descriptions of problems rather than experimental work regarding potential solutions (Ashworth, 2014;Bono et al., 2009;Gabbard et al., 2014;Hyland and Rutigliano, 2013;Saal et al., 2014;Santuzzi et al., 2014).

Search 2: broader psychological research
The results presented in Table 3 show the number of papers mentioning neurominorities and mental health conditions in a sample of UK-based psychology journals.

An empty review of neurodiversity research
The UK-based psychology journal search revealed an increasingly disproportionate body of research focusing on mental health needs and autism, relative to their prevalence. Closer inspection of two journals revealed consistent cross-disciplinary research between mental health and occupational psychology: 1686 studies referenced "work" in the British Journal of Clinical Psychology, and 413 studies on mental health were published in the JOOP, which suggests a more mature research field concerning the impact of mental health in the workplace, though we concede that this has not been quality checked. Conversely, dyslexia studies appear to be decreasing in number and mainly focused on the diagnosis of children to the exclusion of occupational contexts. ADHD, Dyslexia, DCD and TS are all underresearched compared to their population prevalence.
Search 3 and question 4: multi-disciplinary studies on neurodiversity and work Of the 48 extracted papers, the following conditions were represented: ADHD, 13 papers; autism, 19 papers (including four with strong evidence supporting inclusion activities); DCD, five papers; dyslexia, six papers and TS, five papers. Table 4 shows the professional discipline sources (indicated by journal title) of the 48 papers. We note that while neuroscientific research has proliferated in recent years and accounts for much of the extant condition-specific literature, only one neuroscience paper met the criterion of referring to occupational issues. A full list of the journal titles for extracted studies is presented in Appendix.

Synthesis Question 1. What, if anything, can we infer from the current literature about
"what works" to improve the inclusion of neurodivergent individuals at work?
Given the absence of relevant intervention evaluation (either randomized control trial, crosssectional research, process analysis or qualitative) in mainstream organizational and management psychology, we consider our first search an "empty review" (Yaffe et al., 2012). Further, search two indicated that the investigation of individual neurominority conditions per se is unrepresentative of population prevalence and seems instead to be led by media fashion; for example, the dominance of autism within current popular narratives (Bernick, 2019). The limited number of primary papers across broader applied psychological domains describes issues and calls for further research, yet do not culminate in a clear research agenda congruent with evidence-based practice (Schlosser and Sigafoos, 2009). Regarding our third review question concerning quality of evidence for determining "what works", we note limited research that could contribute to evidence-based agenda which is dispersed across disciplines lacking relevance to the socio-legal, economic and occupational context; for example, neuroscience, child development and social work, which leaves crucial   (Chaytor et al., 2006;D'Souza and Karmiloff-Smith, 2017). Vocational rehabilitation and occupational health are the closest represented disciplines in terms of purpose, specifically exploring conditions, their impact on the workplace and intervention evaluation. However, these fields are more closely aligned to the medical paradigm, which implicates a disability inclusion model for practice rather than an organizational-level talent management approach such as the mentoring and peer support programs favored by gender and race inclusion initiatives (Roberson, 2018). We conclude that there is insufficient academic inquiry into neurodiversity at work and highlight a need for broadening the epistemological frameworks of neurodiversity research to offer theoretical support to the developing agenda.
Synthesis Question 2. How should the prospective research agenda be epistemologically framed?
To address the disparity between research to date and the emerging social movement arguing for marginalized lived experience at the heart of research (Bottema-Beutel et al., 2020; Rios et al., 2016), a constructivist investigation of experience and power dynamics appears crucial. Alongside, we recognize the need for positivist evaluation of interventions; the limited extant studies indicate an immature evidence base to underpin practice and policy. Controlled trials must be broadened from the individual to the organizational level of analysis. Minded that qualitiative and quantitative research are complementary, not conflicted, we propose critical realism (Houston, 2014) as an appropriate epistemology for a developing research agenda. Critical realism facilitates inductive, hypothetic-deductive and abductive reasoning, required to iteratively blend theory and method (Van Maanen et al., 2007) for the development of a nascent research field in a pragmatic paradigm (Simpson, 2018). Our proposed ecological framework is outlined in Figure 2, deploying multiple levels of analysis (Bronfenbrenner, 1979) in which we focus on two intersecting points for intervention and analysis: the micro/ meso and the meso/macro. This is congruent with critical realist research and incorporates the ecological approach that has been more thoroughly explored in general disability employment (Szymanski et al., 2012). We now briefly explore the issues within each level of analysis before outlining a research and practice agenda in more detail in the next section. The micro level refers to the neurodivergent employee, their individual work performance and the current perceived need to attain levels of literacy, sedentary concentration and/or social communication norms akin to their neurotypical co-workers in order to retain their employment. Individual level effects will be additionally influenced by intersectionality, such as marginalization resulting from gender, gender identity, ethnicity and race, sexual orientation, age and socioeconomic status (Crenshaw, 1991;Cucina et al., 2013;Greenhaus et al., 1990;Ozbilgin et al., 2011;Szymanski et al., 2012;Young et al., 2020). Critical realism facilitates evaluative studies regarding individual level accommodations as described above, as well as exploring the influences on career inclusion caused by the power dynamics; for example, self-efficacy and career awareness (Leather et al., 2011;Nalavany et al., 2017;Test et al., 2009).
The meso-level refers to the immediate context of the neurodivergent employee, including design of their job, workstation and environment, but also the communication norms of their team and organization. The meso-level is influenced heavily by the socio-historical context of said norms including the prioritization of literacy and fine motor control as the major form of communication, and sedentary, hyper-social education and work (Doyle, 2020). We also note current occupational constructions drawn from the macro-historical level of analysis which EDI create meso-level norms. For example office work, which made sense when staff needed access to paper files and ledgers in the 19th and 20th centuries, but makes less sense now that our work is stored electronically and remotely. Such norms have created exclusion and segregation; for example, the illiterate dyslexic who should be avoided or the autistic technologist who might be desirable but only in a limited role. Thus, alongside a "Realist" evaluation of direct support effectiveness for employers such as toolkits and inclusion best practice guidance, a "Critical" challenge of common/absent narratives and role models in business is encouraged.
The macro-level of analysis influences neurodiversity inclusion through the use of active labor market policies such as those described in the introduction (e.g. Access to Work, Jobs in Jeopardy) and the extent to which employers are incentivized to address disability inclusion in hiring and talent management (Murfitt et al., 2018;Saleh and Bruy ere, 2018). As mentioned in the introduction, such policies are lacking in longitudinal evaluation, which is an urgent concern considering the persistent disability employment gap more broadly. Yet such policies also exist in a macro-historical context of wider technological advancement that facilitates new possibilities for inclusion. Contemporary expectations of normalization and homogeneity at work, inherited from the industrial revolution as above, are in flux as we adapt to ubiquitous technology, cloud access to records and indeed even the type of work available (ONCE, 2019; WEF, 2020), impacting the policies and institutions just as they affect businesses. For neurominorities who have struggled with tasks typically framed as necessary for "good" performance, including literacy and handwriting, hyper-socialization and sedentary work, such changes require critical appraisal as they offer potential for increased inclusion across the lifespan. For instance, writing tasks are already becoming more fully supported through adaptive software, which allow more accurate dictation and transcription; these changes will An empty review of neurodiversity research increasingly be incorporated into education and national training design. We see opportunity here for the next generation of neurominority workers, who could be significantly less disabled if policies can catch up with technological trends. For example, incentivising acceleration of digitization and automation of literacy production in education and public sector workplaces could have a ripple effect to meso-level workplace norms more generally. The next section focuses on specific research activities required to mature the neurodiversity field of practice, building on the critical realist epistemology framework to evaluate and challenge both supply and demand approaches.

Discussion: mapping a new field of research and practice
The research agenda Given the current dearth of evidence as evidenced through our empty review, our attention must turn to the focus for a new research agenda. One particularly urgent concern is the need for longitudinal evaluation and process analysis of adjustments that are currently routinely deployed at the micro-level, such as coaching and assistive technology (Doyle, 2019a, b). We need to understand how accommodations "work" as interventions for a number of target outcomes, such as secure employment, improved career advancement, added value for employers through enhanced work performance, but ultimately evidenced by a reduced disability employment gap. Generalist disability research has produced some cost benefit analysis of supply side adjustments as well as how to implement productive demand side interventions, such as networking, best practice sharing, reporting on disability inclusion data and influencing the will of businesses in acting inclusively (Saleh and Bruy ere, 2018). The neurodiversity movement needs replicate, in order to evidence the business case argument that inclusion equals talent and may improve morale and productivity (ILO, 2010).
Alone, intervention evaluation of which activities deployed in practice "work" will only focus on change at the micro level, with meso and macro "success" markers as an aggregate of individual improvement and achievement. Yet, globally, current legislation requires the employer to make accommodations/adjustments to infrastructure; it does not require the individual employee to change (social model). Without multi-level, critical analysis we risk appropriating medical research agendas that ignore the social model identity of that minority (Huijg, 2020;Riddick, 2001;Shakespeare and Watson, 1997). Such limited perspective may not only fail to recognize individual strength and capabilities but also resonate poorly with an increasingly vocal stakeholder movement who ascribe to the "nothing about us without us" (Charlton, 1998) principle (Bottema-Beutel et al., 2020. We argue that we can learn from principles adapted from analogous fields such as race and gender inclusion, which focus on identifying unconscious biases in hiring, managing and promoting marginalized workers (Roberson, 2018). Outcomes such as pay disparity and representation at senior levels are considered markers of success, not the effectiveness of programs to improve the performance of individuals who happen to be Black, trans or female. Race and gender interventions are thus less medicalized, without the "diagnosis" and "treatment" lens favored in supply side interventions for neurominorities, though we acknowledge that the social "norms" of the White male are still implicitly applied through widespread mentoring arrangements (Dashper, 2019;Kandola, 2018). To summarize, operating from the social model signposts that, like wider equality diversity and inclusion initiatives, it is businesses that need to change not individuals. There is a clear need for neurodiversity research which incorporates a macro-economic, socio-legal perspective and identifies areas within the employee life cycle where policy change would lead to greater inclusion.

EDI
We look to generalist disability inclusion research again for further insight on macroeconomic research directions. Herein, it remains unclear whether demand side intervention policies such as tax breaks, quota targets and penalties stimulate systemic change (Salah and Bruyere, 2018), although there is a moral (and legal) imperative for starting at the system level. Current organizational practice in neurodiversity inclusion, however, tends more toward the "carrot" than the "stick". Key stakeholders in business may not respond positively to mandated quotas, and this could set the movement back rather than drive acceptance. Building on piloted affirmative action programs such as autism at work initiatives, employers may instead prefer to be guided by clear policy suggestions on how to progress from token inclusion to systemic inclusion. Research activities will again need to replicate this sort of analysis to support the development of policy through evaluation of neurodiversity specific endeavors currently in vogue such as affirmative action.
To summarize, mainstream psychology and management research have failed to heed the call for research to support the needs of a significant minority (Beauregard et al., 2018;Ozbilgin et al., 2011), and interventions remain focused at the individual level. We argue that this is short-sighted and propose a series of research questions, presented in Table 5, using the "Context, Intervention. Mechanism, Outcome" (CIMO) framework as a guiding structure in Realist research . Our critical lens acknowledges the ontological duality of medical and social models in neurodiversity and enables both inductive and deductive research, in iterations and cycles rather than a beginning and an end. The critical realist paradigm epistemologically frames the need for research to simultaneously consider "what works" with appropriate challenge to the dominant paradigm of what is "not working."

Practice implications
The neurodiversity paradigm is yet to make an evidence-informed impact on organizational practice. To advance practice whilst awaiting the development of specific research, we propose applying "Universal Design"; a set of principles about justice, flexibility and simplicity, currently aimed at increasing accessibility of buildings and technology (The Center for Universal Design, 1997) to HR processes. Our overarching premise is to ensure inclusive practice throughout the employment lifecycle to pre-empt (rather than remediate) exclusion and bias embedding activities. Examples might include replacing unnecessary interview requirements for jobs that hinge on data analytics (where work sample testing would be more appropriate and tap into the typically autistic talent of accuracy, for example) or considering the overuse of cognitive testing, including without assistive technology compatibility, where specialist thinkers are required to demonstrate generalist abilities for specialist roles. Both these activities could improve the predictive validity of recruitment and benefit a wider population. In short, HR could adapt to maximize the talent potential of this untapped minority, without compromising the integrity of recruitment and organizational performance. Table 6 indicates how universal design principles could be applied across six core areas of HR: job design, hiring, onboarding, training and development, performance management and well-being support. We recommend the review, critique and further development of these ideas in industrial/organizational psychology and human resource management research.
Additionally to universal design, other bodies of work may offer guidance to practitioners. Those with a minority diagnosis are unlikely to disclose to employers (Madaus et al., 2002;Santuzzi et al., 2014), for (warranted) fear of prejudice and stigma (Colella et al., 1998). Employment inclusion practitioners could learn from more mature research bodies investigating minority prejudice such as mental health, race, gender or sexual orientation to develop protocols for increasing equity (Kessler et al., 2009;Roberson, 2018). Practitioners should, however, simultaneously balance the legal obligations for disability support and potential medical treatment that might be in operation at the micro level. In practice, relevant EDI interventions need to include, but go beyond seeking qualified, professional support for individual employees to ensure that relevant activities do not remain merely compliancefocused. More fundamentally, there is a need for practitioners to intervene when delivering standard HR protocols around job design, hiring, contracting, training, management performance and well-being to consider the impact of the neurotypical norm (as per Table 6). For example, does a role really require influencing skills and teamwork if the focus is on highly accurate output produced alone? Should a role require full-time attendance to a busy, crowded office space which has an additional cognitive burden for those with attention deficits, given the ubiquity of remote access technology? HR, applied psychologists, occupational health and vocational rehabilitation practitioners can deliver systemic change by updating their practice to incorporate neurodiversity themes, consulting with lived experience voices to ensure participation and challenging "neuronormativity" (Huijg, 2020). Lived experience voices are essential for empowering participation; we recommend that organizations should facilitate neurodiversity champions to speak to the full variety of neurominority experiences and appreciate the different, and often conflicting, narratives of the people for whom they speak (Baron-Cohen, 2019;Doyle, 2021). However, we note that in-house champions should not be expected to provide professional service voluntarily alongside their paid roles. Accommodation and inclusion policy initiatives should still be advised by occupational professionals with expertise in disability (Spargo-Mabbs et al., 2020). There are harsh legal penalties for failing in practice, both for individuals who may lose their job and employers who could be sued. As neurodiversity inclusion currently sits at the intersection between disability accommodation and diversity inclusion, practitioners must walk a tight rope between the medical and social models to improve occupational access and success. Current advice is therefore to target all levels of analysis in weighing up appropriate interventions, combining participation from those affected with professional recommendations and services. Practitioners are encouraged to seek evaluation opportunities and to work with academic partners on longitudinal studies to build the evidence base in this area as per the activities in Tables 5 and 6. In Table 7, we pose a series of reflective questions for practitioners to consider when beginning to adopt neurodiversity inclusive protocols and projects.

Conclusion
In this paper, we have systematically described the dearth of consistent and ecologically valid research regarding neurodiversity at work. This, we argue, is a risk for individuals, employers, policymakers and applied disability inclusion practice. The neurodiversity phenomenon is no longer new, and we thus present it as an exciting opportunity for multidisciplinary researchers to explore the implications for both individual career interventions and organizational design. We have outlined an epistemological starting point for the field of research to begin answering the pressing question of "what works". The practical concerns we have outlined contribute to a wider need to reflect on the broader influence of disability inclusion legislation which, despite decades of operation in developed countries, has yet to make a substantive impact on disability employment (Scope, 2018;WHO, 2011). We indicate a need to build further than the consideration of "what works", to "what can we make work", by incorporating appropriate structural flexibility to facilitate positive work, life and societal outcomes for the many neurodiverse "diamonds in the rough". We conclude with a clear message for five shifts in thinking as depicted in Figure 3.
We advocate for starting with a move from the medical to social perspective, then embracing context-sensitive and inclusive research and interrogating policy and practice based on legal compliance. Learning from other, more mature fields is a necessary step to enable organizations to move toward inclusive design without repeating mistakes or reinventing the wheel. We hope we have inspired researchers and practitioners alike to critically reflect on the gaps in our understanding and begin building neurodiversity into An empty review of neurodiversity research studies and projects. We will know this is working when we see neurominorities represented in all forms of employment, adding critical diversity of thought and experience to the design, delivery and support of human endeavors. Inclusive work practices benefit everyone.