Assessment Integrity or Linguistic Exclusion? AI, Viva Exams, and Educational Justice

Language is never neutral. The ways we speak, write, and communicate shape not only how others perceive us, but also the opportunities we can access within educational and professional worlds. In the sociology of education, Basil Bernstein’s (1971, 1973) theory of elaborated and restricted codes offers a foundational lens to examine how certain linguistic styles become privileged as markers of intelligence and competence.

Yet in contemporary contexts, the question of language, power, and assessment is complicated further by emerging challenges such as generative AI, the potential rise of viva-style exams to combat it, and their implications for neurodiverse students. This article explores how Bernstein’s codes remain relevant today, how individuals acting as verbal chameleons navigate shifting linguistic demands, and why proposed AI-proof assessments risk reproducing old inequities in new forms.

Drawing together sociological theory, neurodiversity perspectives, and critical reflections on academic integrity, it asks: who benefits when we redefine what counts as valid knowledge, and at what cost?

How Schools Turn Language into Class Advantage

Building on Bernstein’s conceptualisation of elaborated and restricted codes, Pierre Bourdieu (1991) extends this analysis by framing language as a form of cultural and linguistic capital within educational fields. For Bourdieu, elaborated code functions as a marker of distinction, signalling alignment with dominant class-based habitus, aesthetic tastes, and epistemic styles that are implicitly recognised and rewarded within formal education. Schools not only privilege elaborated code; they treat it as neutral, natural, and universal, embedding its valuation within curricula design, pedagogical expectations, assessment structures, and teacher-student interactions. This creates a landscape where linguistic capital is not merely about vocabulary, clarity, or syntax but becomes inseparable from embodied dispositions such as confident oral delivery, ease with abstract theorisation, and the ability to frame knowledge in institutionally preferred ways.

Bernstein’s focus on codes highlighted how elaborated code enables knowledge to be standardised and decontextualised for assessment, fitting within an educational epistemology that equates abstraction with intelligence and communicative value. However, critics argue that Bernstein’s framework underplays the relational and cultural power embedded within elaborated code’s dominance, overlooking how institutional privileging of this code also entails the devaluation of forms of speech grounded in localised meaning, affect, and relational knowledge (Reay, 2004; Heller, 2002). In this way, elaborated code is never simply a neutral linguistic resource; it is a gatekeeping mechanism that reflects and reproduces middle-class cultural hegemony.

Bourdieu’s concept of symbolic violence deepens this critique. When students are told to ‘speak properly,’ it is not a benign pedagogical correction aimed at clarity; it is an enforcement of dominant cultural norms and expectations disguised as neutral educational standards. This process enacts symbolic violence by presenting middle-class linguistic practices as universal benchmarks of intelligence, rationality, and credibility, while simultaneously delegitimising other ways of speaking and knowing. Even when working-class students master elaborated code, they remain structurally marginalised if they do not also embody the aesthetic dispositions, accents, and performative confidence aligned with dominant middle-class habitus (Reay, 2017). Their fluency is treated as learned performance rather than authentic belonging, often subject to greater scrutiny than their middle-class peers.

Furthermore, Bourdieu (1986) emphasises that elaborated code proficiency alone does not guarantee upward mobility. While linguistic capital is necessary for educational success, it is insufficient without accompanying social capital – the networks, relationships, and cultural cues that grant access to elite spaces. This creates a structural paradox: linguistic adaptation is necessary for participation within education, yet it rarely enables full integration into dominant class positions. Students who become ‘verbal chameleons’ by shifting between codes do demonstrate agency, but this agency is exercised within a system that demands adaptation to dominant expectations as a condition of recognition.

Integrating Bernstein and Bourdieu thus reveals that linguistic inequality is not merely about communicative style or vocabulary mastery. It is a structural process whereby language, class, and power intersect to reproduce educational stratification, subtly legitimising middle-class norms as universal standards. The persistent misrecognition of elaborated code as an indicator of innate intelligence, rather than as a product of privileged socialisation and cultural immersion, sustains symbolic violence within educational and professional contexts. This ensures that those already advantaged by birth and upbringing continue to define what counts as knowledge, competence, and worth within institutional hierarchies.

Flexibility, Alienation, and Neurodivergent Exclusion

Leading from Bernstein and Bourdieu’s analysis of linguistic capital, some students can be conceptualised as verbal chameleons – individuals able to shift fluidly between elaborated and restricted codes depending on their audience, context, and purpose. This linguistic flexibility allows them to fulfil the expectations of educational institutions while retaining ties to their community language practices. From a Bourdieuian perspective, such code-switching is not simply an individual skill but reflects strategic deployment of linguistic capital to maximise recognition within structures that privilege elaborated code. Verbal chameleons demonstrate agency within constraint, leveraging their ability to adapt to institutional norms as a mechanism of survival and advancement.

However, this adaptability comes at a cost. While verbal chameleons may navigate institutional demands effectively, they often remain outsiders within elite contexts. Their fluency in elaborated code is read as an acquired performance rather than an authentic expression of belonging, leading to ongoing scrutiny and the risk of being labelled ‘inauthentic’ or ‘trying too hard’. Simultaneously, within their home communities, their use of elaborated code may be perceived as pretension or disloyalty, creating an internal conflict and sense of alienation from both worlds. Thus, their linguistic flexibility, while strategic, does not dismantle structural inequalities but instead highlights the tension between conditional inclusion and continued marginalisation.

In contrast, for many neurodivergent students, code-switching is not a straightforward or even possible task. Some autistic students, for example, may produce highly sophisticated written elaborated code characterised by precision, structured logic, and analytical depth, yet struggle profoundly with spoken elaborated code that requires rapid processing, social-pragmatic nuance, and dynamic rhetorical shifts. Others may thrive in oral, context-bound communication where shared meaning is assumed but find the linear, structured demands of academic writing restrictive or meaningless. Education systems rarely recognise these communicative differences, often framing them as deficits or weaknesses rather than forms of cognitive diversity with equal epistemic value.

Furthermore, neurodivergent students often face an additional layer of symbolic violence. While verbal chameleons are celebrated for their flexibility (even if only conditionally accepted), neurodivergent students whose styles do not align with elaborated code norms are marginalised, excluded, or pathologised. Their communicative styles are evaluated against neurotypical, middle-class linguistic expectations, embedding ableism within educational language hierarchies. This reproduces structural inequality by denying legitimacy to alternative ways of knowing, processing, and expressing knowledge.

This juxtaposition thus reveals a profound sociological dichotomy: verbal chameleons are praised for their strategic adaptation to dominant codes, despite experiencing alienation, while neurodivergent students are penalised for their divergence from these norms, regardless of their analytical capacity or conceptual understanding. Both groups expose how elaborated code functions as a gatekeeping mechanism that privileges communicative conformity over cognitive depth, shaping the conditions of inclusion, exclusion, and the reproduction of power within educational fields.

Will Oral Exams Solve AI Cheating – Or Create New Inequalities?

As generative AI technologies rapidly disrupt the landscape of student assessment, universities are increasingly concerned about the integrity of written assignments that can now be outsourced, paraphrased, or fabricated with startling ease. In response, some institutions have begun revisiting the idea of viva-style oral examinations — an assessment model that, on the surface, seems to promise a robust safeguard against AI misuse. Since a viva requires students to articulate their thinking in real time, drawing on memory, synthesis, and spontaneous analysis, it is assumed that students cannot simply present AI-generated work as their own when asked to explain, defend, or extend it verbally.

However, this proposed solution demands closer sociological scrutiny. From a Bernsteinian perspective, viva exams privilege the mastery of elaborated code in its most performative form: real-time, decontextualised, precise, and fluent. Students must demonstrate not only subject knowledge but the ability to produce polished academic discourse under pressure — a linguistic demand deeply entangled with classed patterns of socialisation. Middle-class students, whose home environments often nurture elaborated code, tend to enter higher education already prepared for this style of expression. For them, oral defence feels more like a demonstration of what they already know how to do.

For some disciplines and professional contexts, viva-style assessment has long been used effectively to test applied knowledge, oral communication skills, and the ability to think on one’s feet. In fields such as medicine, law, or counselling, the capacity to explain, synthesise, and defend decisions verbally under pressure can be directly relevant to future practice.

By contrast, working-class students and those from non-dominant linguistic backgrounds may possess equal — or even greater — conceptual knowledge but lack the ease or confidence to produce polished spoken elaborated code under the artificial time constraints of a viva. Here, Bourdieu’s concept of symbolic violence becomes visible: the spoken performance of knowledge is misrecognised as a neutral measure of intellectual worth when it is, in fact, a reflection of social and cultural capital (Bourdieu, 1991). The viva, like other forms of assessment, thus risks reproducing the very class-based gatekeeping it claims to circumvent.

Verbal chameleons — those who can shift fluidly between codes — may fare better under such regimes. Their adaptive linguistic capital allows them to perform the required fluency convincingly, even if they privately experience anxiety or alienation from both their communities of origin and the dominant academic sphere. Yet, as this article argues, even their celebrated adaptability is conditional. It does not challenge the structural bias that equates elaborated, fluent spoken discourse with merit.

For neurodivergent students, the implications are starker still. Many autistic learners, for instance, excel when given the space to compose written work that is precise, rigorous, and analytically rich — often exceeding their neurotypical peers in depth of thought and originality. Yet when required to reproduce this knowledge verbally, they may struggle with real-time processing, sensory stress, or social-pragmatic demands that bear little relation to the quality of their intellectual labour. Here, the viva becomes not an assurance of academic integrity but an instrument of epistemic exclusion, favouring students whose communicative profiles match the unspoken norm.

Furthermore, the structural realities of viva-style assessment are often overlooked in the rush to ‘AI-proof’ higher education. Conducting rigorous oral examinations at scale requires significant staff time, consistent examiner training, and robust mechanisms to ensure fairness and consistency. In a sector already under immense resource pressure — with larger cohorts, marketised funding models, and precarious staffing — adding thousands of individual oral exams risks creating new inequities. It may shift the burden onto overworked faculty or produce inconsistent grading when subjective impressions of fluency are mistaken for indicators of critical capacity.

A critical realist lens (Bhaskar, 1979) helps situate this dilemma: while the structural risk of AI undermining written work is real, so too are the deeper structures of classed and ableist bias that underpin the valorisation of spoken elaborated code. Without addressing these structures, a surface-level reform such as the viva risks merely reproducing old hierarchies in a new guise.

The question, then, is not simply whether oral exams can outwit AI, but what hidden assumptions about valid knowledge, valued communication, and deserved success they embed. Do they test what we claim to value — conceptual mastery, creativity, and critical synthesis — or do they reward confidence, fluency, and cultural alignment? What of the student whose brilliance lies in patient, written construction of ideas but who falters when asked to improvise under the bright lights of a one-to-one oral exam?

This is not to say that viva-style assessment should be rejected outright. In certain contexts — for example, doctoral defences or professional practice training — it can play an important role in verifying authentic engagement. But as a blanket solution to the AI challenge, its ethical and sociological implications must be confronted openly. Educators and policymakers must ask whether they are addressing technological threats to integrity by inadvertently deepening existing exclusions.

In this sense, the viva is not just a pedagogical tool but a mirror reflecting the educational system’s entanglement with the reproduction of privilege. Unless accompanied by genuine structural reforms — flexible, inclusive assessment design; recognition of neurodiverse communicative profiles; and a commitment to de-centring elaborated code as the sole marker of competence — oral exams risk solving one integrity problem while entrenching deeper injustices.

Towards Linguistic Justice in an AI Age

The shifting landscape of higher education, challenged by the rise of AI and its implications for assessment, reveals persistent truths about language and power. From Bernstein’s codes to Bourdieu’s symbolic violence, verbal chameleons to neurodivergent communicative exclusion, the sociology of language illuminates how education continues to privilege those whose linguistic and cognitive styles match institutional norms. Viva-style exams, framed as a solution to technological threats, risk entrenching these inequities by privileging performative fluency over conceptual depth and favouring students whose cultural, class, or neurotypical habitus aligns with dominant expectations.

As universities adapt to the realities of AI, they face a choice: to reinforce existing hierarchies under the guise of integrity, or to rethink assessment in ways that value diverse ways of knowing, processing, and communicating. Ensuring rigour need not come at the cost of inclusion. True educational justice requires challenging assumptions about what counts as knowledge, who decides, and whose voices remain unheard in the pursuit of innovation.