Ableism by Design? Artificial Intelligence,Special Education Law,and Policy Implications for Equity in P–16 Education

Abstract

This qualitative study examines how artificial intelligence functions as a policy actor within special education systems in local school districts across P to 16 education. Guided by Disability Critical Race Theory and Critical Policy Analysis, the study draws on interviews with special education professionals and analysis of institutional documents, including district guidance and internal protocols. Findings show that AI is framed as neutral and efficient, yet it reproduces ableist and racialized norms, constrains professional judgment, and operates amid policy silence. The study underscores the need for equity centered governance of AI in special education.

Introduction

Artificial intelligence technologies are being rapidly integrated across P to 16 educational systems, reshaping instructional delivery, assessment practices, data management, and decision-making processes. In kindergarten through grade 12 contexts, AI powered tools are used for early screening, progress monitoring, behavioral tracking, and individualized learning platforms, while in postsecondary settings, algorithmic systems manage disability accommodations, predict academic risk, and monitor student engagement (Holmes et al., 2019; Selwyn, 2019). These developments are often framed as innovative solutions to challenges related to efficiency, personalization, and accountability. However, as AI becomes embedded within educational infrastructure, its implications for special education policy and equity demand sustained examination, particularly given its growing influence on high stakes decisions.

Within dominant educational policy discourse, artificial intelligence is frequently positioned as neutral, objective, and efficient, with the potential to reduce human bias and enhance fairness (Williamson & Eynon, 2020). These framings align with technocratic narratives that privilege data driven decision-making and standardization. Yet scholars have demonstrated that AI systems are socially constructed, reflecting historical data, institutional priorities, and dominant cultural norms (Benjamin, 2019; Noble, 2018). In special education contexts, these systems contribute to the construction of disability by reinforcing normative expectations around learning, behavior, communication, and productivity.

Despite the expansion of artificial intelligence, special education policy has not evolved to address its implications. Foundational civil rights statutes such as the Individuals with Disabilities Education Act, Section 504 of the Rehabilitation Act, and the Americans with Disabilities Act do not account for algorithmic governance or automated decision-making. As a result, there is limited guidance on how AI tools should be evaluated, regulated, or constrained when they influence decisions related to disability identification, placement, and services, creating ambiguity around accountability and due process (Ifenthaler et al., 2024). Moreover, AI adoption often occurs through administrative decision-making rather than democratic policy processes, driven by vendor claims and institutional pressures tied to compliance, staffing, and fiscal constraints (Hillman, 2023). These conditions position artificial intelligence as a powerful policy actor that shapes how disability is governed across P to 16 education, often reinforcing structural inequities while remaining largely invisible within formal regulatory frameworks.

Purpose of the Study

The purpose of this study is to examine how artificial intelligence functions as a policy actor within special education systems and to explore the implications for equity across P to 16 education. Drawing on the perspectives of special education administrators, IEP decision makers, and postsecondary disability services professionals, this study investigates how AI is understood, implemented, and justified within existing educational policy frameworks. By foregrounding disability, race, and power, the study seeks to illuminate how AI technologies may reproduce or disrupt ableist and racialized assumptions embedded in special education governance. The following research questions guide this inquiry:

1. How do special education professionals understand and experience artificial intelligence within existing educational policy frameworks

2. In what ways do artificial intelligence systems reproduce or challenge ableist and racialized assumptions in special education decision making

3. What policy implications emerge for advancing equity for students with disabilities across P to 16 education

Addressing these questions is critical at a moment when educational systems increasingly rely on AI driven tools to inform decisions that have profound consequences for students with disabilities. By centering the perspectives of professionals who enact special education policy in practice, this study contributes to ongoing debates about technology, governance, and civil rights in education. The findings aim to inform more intentional and justice-oriented AI policymaking that aligns technological innovation with the foundational goals of equity, inclusion, and educational opportunity for all learners.

Theoretical Framework

Disability Critical Race Theory (DisCrit) as a Framework for Educational Policy Analysis

Disability Critical Race Theory, commonly referred to as DisCrit, provides a critical framework for examining how disability and race function as interlocking systems of power within educational policy and practice. Emerging from the convergence of critical race theory and disability studies, DisCrit rejects deficit-oriented understandings of difference and instead interrogates how institutional structures produce and sustain inequity (Annamma et al., 2013; Sabnis & Bueno Martinez, 2021). Central to this framework is the assertion that racism and ableism are mutually constitutive forces that shape how students are categorized, governed, and valued within schools and universities. This lens is particularly relevant for analyzing artificial intelligence in education, where claims of neutrality and efficiency often obscure the racialized and ableist assumptions embedded within technological systems.

A core tenet of DisCrit is the critique of normalcy as a socially constructed and policy enforced standard. Educational policies rely on normative benchmarks related to cognition, behavior, communication, and productivity to determine access to services and inclusion, often presenting these benchmarks as objective despite their racialized and classed foundations (Annamma et al., 2013). DisCrit challenges these assumptions by foregrounding how such standards function as mechanisms of exclusion, particularly for students at the intersection of racial marginalization and disability identification. In the context of artificial intelligence, these benchmarks are frequently encoded into algorithms, transforming historically contingent norms into automated decision-making criteria. DisCrit also centers the lived experiences and counter narratives of marginalized communities as essential for understanding policy impact, emphasizing how students and professionals experience and resist systems in ways that diverge from official intentions (Connor et al., 2016; Ellis, 2025).

Applying DisCrit and Critical Policy Analysis to Artificial Intelligence in Education

Disability Critical Race Theory and Critical Policy Analysis together provide a critical framework for examining how educational policy and technology function as mechanisms of power rather than neutral tools. DisCrit foregrounds the interlocking nature of ableism and racism, while Critical Policy Analysis emphasizes that policy is socially constructed, politically contested, and enacted through uneven relations of power (Ball, 1994; Diem et al., 2014; Ozga, 2000). From this perspective, policies and technologies framed around objectivity, accountability, and standardization actively shape which bodies, behaviors, and ways of knowing are deemed legitimate, often naturalizing inequities by attributing disparities to individual deficits rather than structural conditions (Connor & Ferri, 2010; Ellis & Flake, 2026; Ladson Billings, 1998).

The integration of artificial intelligence into special education intensifies these dynamics. AI tools function as policy technologies that carry normative assumptions and enact institutional priorities, even in the absence of explicit regulation (Hillman, 2023; Shore & Wright, 2011). Frequently adopted to enhance compliance, reporting, and efficiency, these systems extend forms of surveillance and control that disproportionately affect students with disabilities and students from racially marginalized communities (Benjamin, 2019; Eubanks, 2019). Because AI systems rely on historical datasets shaped by inequitable schooling practices, they transform past injustices into future predictions, reinforcing biased disciplinary patterns, unequal access to resources, and disproportionate identification (Benjamin, 2019; Noble, 2018). Through a DisCrit lens, these data driven processes are not neutral reflections of reality but active reproductions of inequality.

AI embedded data systems privilege narrow definitions of success aligned with speed, compliance, and standardization, reflecting ableist assumptions about learning and productivity. These values disadvantage students whose communication, behavior, or learning trajectories diverge from normative expectations. As such, artificial intelligence operates as a policy technology that can reify deficit-based constructions of disability and difference, shaping eligibility decisions, placement recommendations, and accommodation determinations across educational contexts. Applying DisCrit and Critical Policy Analysis makes visible how AI both reflects and reinforces structural inequities, underscoring the need for policy frameworks that interrogate power, representation, and justice in educational technology.

Policy Context: Special Education and Artificial Intelligence Governance in the United States

Foundations of Special Education and Civil Rights Policy

Special education policy in the United States is grounded in a civil rights framework intended to protect students with disabilities from exclusion, discrimination, and denial of educational opportunity. Central federal statutes include the Individuals with Disabilities Education Act, Section 504 of the Rehabilitation Act, and the Americans with Disabilities Act. Together, these policies establish procedural safeguards related to identification, evaluation, placement, and access, emphasizing individualized decision making, parental participation, and due process. These protections reflect decades of advocacy aimed at ensuring that disability is not used as a justification for segregation or diminished educational opportunity (Yell, 2025). While these policies provide important legal protections, they were developed in an era that did not anticipate the rise of algorithmic systems in educational governance. AI operates within existing legal protections but may reshape how those protections are enacted in practice, particularly through issues of opacity, standardization, and reliance on system-generated outputs.

Policy Assumptions of Objectivity and Standardization

Educational policy has increasingly prioritized objectivity, standardization, and accountability as indicators of effectiveness and quality. These priorities are reflected in assessment regimes, performance metrics, and evidence-based policy mandates that frame data as neutral indicators of educational value (Biesta, 2010). Artificial intelligence aligns closely with these policy logics, offering scalable tools that promise consistency, predictive insight, and efficiency. Consequently, AI is often positioned as a technical enhancement to existing policy structures rather than as a transformation requiring independent scrutiny.

Critical scholars have argued that policies grounded in standardization obscure structural inequities by treating disparities as technical problems rather than manifestations of racism and ableism (Ladson Billings, 1998). In special education, standardized indicators frequently privilege narrow definitions of academic and behavioral success, disadvantaging students whose learning trajectories diverge from normative expectations. When these indicators are embedded within AI systems, normative policy assumptions about ability and productivity are encoded into automated processes, making inequity more difficult to detect and contest (Benjamin, 2019).

Absence of Explicit Artificial Intelligence Guidance in Special Education Policy

Despite the rapid expansion of artificial intelligence in education, there remains limited explicit federal or state level guidance addressing its use within special education. Existing policy documents rarely specify how AI tools should be evaluated for bias, accessibility, or alignment with disability rights law. Instead, responsibility for adoption and oversight is often delegated to districts, institutions, or private vendors, resulting in uneven accountability and fragmented governance (Williamson & Hogan, 2020). This absence of clear guidance is especially concerning in contexts where AI influences high stakes decisions such as disability identification, placement, service allocation, or accommodation approval. Without transparency regarding algorithmic design, data sources, and decision rules, families and advocates face significant barriers to understanding or challenging AI informed outcomes. The lack of regulatory clarity also complicates enforcement of due process protections guaranteed under special education and civil rights law (Pasquale, 2015).

Tensions Between Innovation and Civil Rights Protections

A defining tension within the current policy landscape lies between the push for technological innovation and the obligation to uphold civil rights protections. Educational leaders are encouraged to adopt AI tools as symbols of modernization and efficiency, while remaining legally responsible for ensuring nondiscriminatory practice. When AI systems influence decision making without clear regulatory boundaries, this tension becomes increasingly difficult to manage (Selwyn, 2019). Special education professionals are often positioned as intermediaries who must reconcile AI generated insights with legal requirements for individualized consideration and equitable treatment. In the absence of explicit policy direction, responsibility for mitigating harm is shifted from systems to practitioners. This dynamic mirrors long standing patterns in special education policy where equity failures are attributed to implementation rather than to structural design (Harry & Klingner, 2022).

Policy Implications for Equity Across in P to 16 Education

The absence of coherent artificial intelligence governance in special education has implications on students with disabilities. Early decisions regarding identification, placement, and monitoring shape long term educational trajectories, influencing access to advanced coursework, disciplinary outcomes, graduation rates, and postsecondary opportunities. In higher education, AI supported accommodation platforms and learning analytics continue these patterns, often emphasizing efficiency and risk management over student autonomy and self-determination (Friedman & Nissenbaum, 1996). Viewed through a DisCrit lens, these processes are not isolated. Without intentional policy intervention, artificial intelligence risks becoming another mechanism through which racialized and ableist inequities are reproduced across educational transitions. Addressing this challenge requires policy frameworks that explicitly integrate artificial intelligence governance with the foundational principles of special education and civil rights law.

Methods

Research Design and Epistemological Orientation

This study employs a qualitative research design informed by Disability Critical Race Theory and critical policy analysis to examine how artificial intelligence functions as a policy actor within special education systems. It investigates how AI is interpreted, enacted, and negotiated by professionals responsible for implementing special education policy across institutional contexts, recognizing that policy processes are dynamic, contested, and shaped by power rather than solely by formal regulation (Ball, 1994; Shore & Wright, 2011). Epistemologically, the study is grounded in the view that knowledge about educational policy is socially constructed, politically situated, and historically contingent, with policies understood not as neutral texts but as social practices enacted within institutional constraints and accountability regimes (Ball et al., 2012; Ozga, 2000). Within this framework, artificial intelligence is conceptualized not merely as a technical tool but as a sociopolitical mechanism that structures decision making, normalizes particular forms of knowledge, and produces material consequences for students with disabilities, aligning with DisCrit’s emphasis on the role of power, race, and ableism in shaping educational structures and outcomes (Annamma et al., 2013).

Participants and Recruitment Strategies

Participants were recruited using purposive and snowball sampling to capture the perspectives of professionals situated at critical points of special education policy and practice. Purposive sampling was used to identify individuals with direct experience overseeing, implementing, or responding to artificial intelligence tools in special education contexts (Patton, 2014). Snowball sampling was subsequently employed to extend recruitment by inviting participants to recommend colleagues whose professional roles offered additional insight into artificial intelligence related decision making and policy implementation (Noy, 2008).

Three participant groups were included to reflect variation in institutional context and policy responsibility. Consistent with the study’s use of purposive and snowball sampling, participants were drawn from two states in the southeastern United States, specifically Virginia and North Carolina, where artificial intelligence tools were actively integrated into special education related decision making. Participants represented a mix of suburban and urban districts, as well as postsecondary institutions serving diverse student populations. Although the sample does not include participants from all regions of the United States and does not explicitly represent rural districts, it was intentionally designed to capture depth of experience across key professional roles involved in policy enactment.

The first group consisted of special education administrators and policy implementers at the district or state level who were responsible for compliance, oversight, and technology related decision making. The second group included IEP decision makers and evaluation professionals, such as school psychologists and special education coordinators, who engaged in eligibility determination, assessment, and placement processes influenced by data systems. The third group comprised postsecondary disability services professionals responsible for accommodation systems and compliance with federal disability law in higher education. Together, these groups provided insight into how artificial intelligence is interpreted and enacted across different levels of educational governance.

This sampling approach prioritizes depth and variation in professional perspective over broad geographic generalizability. As such, findings should be interpreted within the context of the study’s regional scope while offering analytic insights that may be transferable to similar policy environments.

Data Collection Procedures

Data were collected through semi structured interviews with participants from each group. Interviews focused on participants’ experiences with artificial intelligence tools used in special education related decision making, their understanding of existing policy guidance or policy silence, and their perceptions of equity, bias, and accountability. Interview questions were designed to elicit detailed narratives about how AI shaped professional judgment, institutional routines, and interpretations of compliance and responsibility (Kvale, 1996).

In addition to interviews, relevant policy and institutional documents were collected when available, including district guidance, procurement materials, and internal protocols related to artificial intelligence or data systems. Document analysis served to contextualize participant accounts and to examine the alignment or dissonance between formal policy language and everyday practice (Bowen, 2009). Data collection procedures were informed by critical qualitative traditions that emphasize reflexivity, power awareness, and ethical responsibility in research relationships (Madison, 2019).

Data Analysis

Data analysis followed an iterative thematic process informed by Disability Critical Race Theory and critical policy analysis. Interview transcripts and documents were read multiple times to identify recurring patterns related to normalization, surveillance, decision making authority, and accountability. Initial coding focused on descriptive categories drawn from participants’ accounts of artificial intelligence use, policy ambiguity, and professional constraint. Subsequent rounds of analysis engaged in more interpretive coding to examine how these patterns reflected broader systems of ableism and racism embedded in educational policy and technological infrastructures (Saldaña, 2025).

Rather than treating themes as purely inductive, analysis was theoretically guided by DisCrit constructs related to normalcy, deficit framing, and institutional control. Particular attention was given to how participants described who benefited from AI systems, who was positioned as inefficient or problematic, and how responsibility for inequitable outcomes was assigned or displaced. This analytic approach enabled examination of artificial intelligence as both a technical system and a policy actor that reshapes governance in special education (Diem et al., 2014; Hillman, 2023).

Researcher Positionality and Reflexivity

Consistent with Disability Critical Race Theory and critical qualitative methodologies, researcher positionality was treated as integral to the research process rather than as a source of bias to be eliminated. The first author’s background in special education, disability studies, and educational policy informed the framing of research questions, engagement with participants, and interpretation of findings. Reflexive memo writing was used throughout data collection and analysis to interrogate assumptions, emotional responses, and analytic decisions (Tracy, 2010). This reflexive stance was especially important given the power dynamics inherent in policy research. Participants often navigated institutional pressures that constrained their ability to critique policy or technology openly. Reflexivity allowed the researcher to attend to silences, hesitations, and contradictions in participant narratives as analytically meaningful rather than as methodological limitations (Pillow, 2003).

Trustworthiness and Ethical Considerations

Multiple strategies were employed to enhance trustworthiness, including triangulation across participant groups and document sources, prolonged engagement with the data, and analytic memo writing (Lincoln & Guba, 1985). Credibility was supported through careful alignment between findings, participant accounts, and theoretical constructs. Transferability was strengthened through thick description of institutional contexts, professional roles, and decision-making processes.

Ethical considerations included informed consent, confidentiality, and protection of participants’ professional identities. All participant names and identifying details were changed, and pseudonyms are used throughout this study to safeguard confidentiality and minimize potential risk. Given the sensitive nature of discussing policy gaps, technology adoption, and institutional accountability, additional care was taken to remove contextual details that could reveal participants’ identities or organizations. The study was guided by an ethic of care consistent with Disability Critical Race Theory, prioritizing respect for participants and a commitment to equity-oriented scholarship (Ellis, 2025; Smith, 2021).

Findings

This study examined how artificial intelligence is understood, enacted, and negotiated within special education systems by professionals positioned at key points of policy implementation. Drawing on interviews with six participants and analysis of institutional documents, including district guidance, procurement materials, and internal protocols, the findings show how AI functions as a policy actor in practice. Across roles, participants described AI as embedded within data systems and decision-making processes rather than as a clearly defined policy initiative. District leaders emphasized efficiency, compliance, and risk management, a framing reinforced in procurement and guidance documents that prioritized data integration and accountability while offering limited attention to disability equity or due process.

At the school and postsecondary levels, professionals described how AI influence decisions related to evaluation, placement, and accommodation processes, often privileging algorithmically generated indicators over contextual judgment. Internal protocols positioned these indicators as default reference points, constraining professional discretion despite formal commitments to individualized decision making. Across interviews and documents, AI systems were treated as technical enhancements rather than policy interventions, contributing to ambiguity around accountability and equity. Analysis revealed four interrelated themes: AI adoption driven by efficiency and compliance, the elevation of algorithmic authority, the reproduction of ableist and racialized norms, and the absence of clear governance structures. These findings demonstrate how artificial intelligence influence human decision making within special education systems, which has the potential to impact education policy.

Theme 1: Artificial Intelligence Adoption Driven by Efficiency, Compliance, and Institutional Protection

Across participant roles and educational contexts, artificial intelligence was consistently described as a tool adopted to respond to institutional pressures related to efficiency, compliance, and risk management rather than as a mechanism explicitly designed to advance equity for students with disabilities. This framing was reinforced through document analysis of district guidance, procurement materials, and internal protocols, which emphasized system wide consistency, reporting capacity, and alignment with accountability requirements. In these documents, AI tools were positioned as technical solutions to administrative demands, with limited attention to disability equity, racialized impact, or civil rights protections. Together, participant narratives and institutional documents illustrate how AI entered special education systems through managerial pathways rather than through policy frameworks grounded in justice.

Paula, a district level special education director in Virginia, explained that AI driven systems were appealing to district leadership because they created the appearance of consistency and defensibility. She noted,

What leadership likes is that the system looks the same everywhere. If we get questioned, we can say this is the tool we use across the district, that everyone is following the same process and using the same criteria. It gives leadership something concrete to point to. When complaints come in or when we are audited, we can say we relied on the system and not on individual opinions. It feels safer because it reduces pushback and limits how much our decisions can be challenged.

Although these tools were often presented as supports for educators, Paula emphasized that their value was closely tied to compliance. As she stated, “It is less about whether it helps the child and more about whether we can show we followed the process.” Jane, a district level special education director in North Carolina, similarly described AI adoption as a response to staffing shortages and workload pressures. She explained,

We are expected to move cases faster with fewer people, and AI is sold as the answer to that problem. The message is that technology will make up for staffing shortages and growing caseloads without changing expectations. It starts to feel like speed matters more than doing the work carefully or really understanding what students need.

Jane noted that conversations about equity or disability rights rarely shaped these decisions. Instead, she observed, “The focus is always on speed and efficiency. Nobody is asking who this actually works for or who it might hurt.” At the school level, Carlos, a school psychologist in Virginia, described how AI informed assessment systems shaped evaluation practices in ways that privileged throughput over professional judgment. He explained,

The system flags students and tells you who needs to move next. It creates pressure to trust the data because that is what administration wants to see. Even when my professional judgment says to slow down or look more closely, it becomes harder to push back against what the numbers are telling us to do.

Carlos expressed concern that such tools narrowed how student needs were interpreted, particularly for students whose profiles were complex. As he noted, “Once the data says something, it becomes harder to slow down and say this child does not fit the pattern.”

Mary, a special education coordinator in Virginia, described AI as embedded within compliance driven infrastructures that emphasized documentation and reporting over relational decision making. She explained, “The platforms are great for audits. Everything is tracked, everything is documented.” However, she also noted that these systems reshaped practice in troubling ways. “What gets measured is what gets prioritized. If it does not fit neatly into the system, it becomes invisible,” Mary explained. From her perspective, efficiency often came at the expense of individualized understanding.

Postsecondary participants described similar patterns at the P to 16 level. Lisa, a postsecondary transition director in North Carolina, explained that AI supported accommodation systems were designed to manage volume and institutional liability. She stated,

The system flags students and tells you who needs to move next. It creates pressure to trust the data because that is what administration wants to see. Even when my professional judgment says to slow down or look more closely, it becomes harder to push back against what the numbers are telling us to do.

Lisa expressed concern that this logic conflicted with the principles of individualized access. “Disability is not one size fits all, but the system pushes it in that direction,” she said. Michael, a postsecondary transition director in Virginia, described how AI driven monitoring tools framed students with disabilities as risks to be managed rather than learners to be supported. He explained,

The system is watching who might fail, who might not persist. Students with disabilities show up on those dashboards a lot. Once they are flagged, they become the focus of increased monitoring rather than additional support. It can feel like the system expects problems instead of recognizing potential.

Michael noted that adoption decisions were rarely framed around equity. “It is always about protecting the institution and meeting metrics, not about whether the tool is actually fair,” he stated.

Viewed through a DisCrit lens, these accounts reveal how artificial intelligence adoption in special education is shaped by policy logics that privilege efficiency, standardization, and institutional protection. By framing AI as a compliance solution rather than as a civil rights concern, educational systems reproduce ableist assumptions about productivity and normalcy while obscuring racialized inequities embedded in data driven decision making. This theme illustrates how AI operates as a policy actor that reinforces existing power structures, even as it is presented as a neutral and objective innovation.

Theme 2: Algorithmic Authority and the Marginalization of Professional Judgment

Across participant accounts, artificial intelligence systems were described as acquiring a form of authority that increasingly shaped decision making in special education, often at the expense of professional expertise. This dynamic was reinforced by institutional documents, including a district level internal protocol for evaluation and referral review, which instructed teams to prioritize system generated indicators when determining next steps and provided limited guidance for documenting professional disagreement with algorithmic outputs. While AI tools were formally presented as supports intended to inform practice, both the document language and participant narratives revealed how algorithmically generated data came to be treated as definitive, narrowing the range of acceptable interpretations and constraining opportunities for contextual judgment. This shift reflected a broader policy logic in which data driven outputs were privileged over professional knowledge, relational insight, and lived experience. Carlos described how algorithmic systems influenced evaluation practices by establishing an implicit hierarchy of evidence. He stated,

The system flags students and tells you who needs to move next. It creates pressure to trust the data because that is what administration wants to see. Even when my professional judgment says to slow down or look more closely, it becomes harder to push back against what the numbers are telling us to do.

For Carlos, algorithmic authority did not eliminate professional discretion outright but made it increasingly difficult to exercise, particularly when data outputs aligned with administrative expectations. Mary similarly described how AI structured IEP processes by determining what information mattered most. She noted,

Once the data is in the system, that becomes the story everyone focuses on. It starts to shape how people talk about the student and what they think is possible. Other information like family context, past experiences, or what we see day to day in the classroom gets pushed aside. If it does not show up in the data, it feels like it does not count.

Mary explained that professional insights about family context, student trauma, or instructional history were often sidelined because they were not easily captured by the platform. As she stated, “If it is not in the system, it is treated like it does not exist,” underscoring how algorithmic frameworks narrowed how student need was understood. District leaders also acknowledged the growing authority of AI systems. Jane reflected on how reliance on algorithmic outputs reshaped leadership expectations. She explained,

When there is data from the system, it becomes the answer people want. If you challenge it, you are seen as subjective or resistant. It shifts the conversation away from professional judgment and toward defending the numbers. Over time, it becomes easier to go along with what the system says than to keep pushing back.

Jane noted that this dynamic discouraged critical engagement with AI outputs and reinforced a culture where professional judgment was expected to align with what the system produced. At the postsecondary level, participants described similar patterns of deference to algorithmic indicators. Michael mentioned that AI driven dashboards increasingly guided conversations about student persistence and readiness. He further added,

The system is watching who might fail, who might not persist. Students with disabilities show up on those dashboards a lot. Once they are flagged, they become the focus of monitoring instead of support.

Michael emphasized that these systems shaped how students were perceived, often framing disability as risk rather than as a dimension of diversity requiring accommodation. Lisa, a postsecondary transition director in North Carolina, raised concerns about how algorithmic authority limited flexibility in accommodation decisions. She explained,

The system sets expectations for what is reasonable, and it becomes harder to argue for anything outside of that. Once those expectations are in place, they start to feel fixed, even when they do not fully fit a student’s situation. You can explain the context and the need for flexibility, but the system still pushes you back to its categories. Over time, it feels like professional judgment has to bend to what the system will allow.

According to Lisa, professional expertise and student narratives were often treated as secondary to what the system defined as acceptable. This shift, she noted, altered the balance of power in decision making, placing greater weight on automated outputs than on human judgment.

Rather than functioning as neutral supports, artificial intelligence systems emerged as authoritative reference points that quietly restructured decision making in special education. Participants’ accounts show how algorithmic outputs came to delimit what could be questioned, defended, or revised, often placing professional expertise in a reactive position. In privileging forms of knowledge that are legible to data systems, these technologies narrowed how student needs were interpreted and addressed. This theme underscores how artificial intelligence shifts the balance of power in special education governance by elevating standardized data over contextual understanding and lived professional insight.

Theme 3: Reinscribing Ableist and Racialized Norms Through Data

Across participant accounts, artificial intelligence systems were described as relying on data models that reflected narrow and historically rooted assumptions about learning, behavior, and success. Rather than disrupting inequitable patterns in special education, these systems often reproduced and intensified them by translating existing disparities into data driven indicators of risk or deficiency. Participants emphasized that students with disabilities, particularly those from racially marginalized backgrounds, were disproportionately flagged, monitored, or categorized in ways that echoed longstanding practices of surveillance and exclusion.

Paula described repeatedly noticing the same patterns emerge in AI generated reports. She explained that “the same groups of students keep showing up on the dashboards,” and clarified that these were most often students of color and students with disabilities. What concerned her most was not the pattern itself but the lack of interrogation it prompted. Because the data was treated as objective, she noted,

People assume it is just reflecting reality instead of asking why the same students are always being flagged. Once the data is presented that way, it feels unquestionable. There is very little conversation about how the system was built or whose experiences shaped it. The focus stays on the numbers, not on the patterns or the inequities behind them.

In this way, historical inequities were rendered natural through data rather than understood as products of policy and practice. Patterns of overrepresentation, surveillance, and exclusion were reframed as neutral outcomes of algorithmic analysis instead of as the result of long-standing decisions about curriculum, discipline, assessment, and access. By presenting disparities as data driven facts, the system obscured the role of institutional structures that have historically marginalized students with disabilities and students from racially marginalized communities. What appeared as objective evidence thus functioned to depoliticize inequality, shifting attention away from systemic responsibility and toward the presumed deficits of individual students.

Jane similarly reflected on how AI systems privileged particular forms of behavior and compliance that aligned with dominant norms. She observed that the systems rewarded students who moved quickly and quietly through educational processes, while those who required flexibility or additional time were more likely to be marked as concerns. According to Jane, this disproportionately affected students with disabilities and students from marginalized communities, not because of inherent deficit, but because:

The system is not built with them in mind. It assumes a very specific way of learning, behaving, and moving through school. When students do not match that model, the system treats them as problems to be managed rather than learners to be supported. There is very little room for flexibility or difference. Over time, it becomes clear that the design reflects who the system was really created for and who it leaves out.

Mary described how data systems compounded deficit-based narratives once a student carried a disability label. She explained that “everything a student does gets filtered through that label,” particularly when behavior, attendance, and academic indicators were aggregated in a single platform. Over time, these patterns came to be interpreted as evidence of struggle rather than as signals of unmet need. As Mary noted,

It starts to look like proof, even when what we are seeing is the system reacting to difference instead of need. The data gets treated as confirmation that something is wrong with the student. That makes it harder to step back and ask whether the system itself is creating the problem.

Carlos reflected on how artificial intelligence tools framed deviation from narrowly defined norms as automatic cause for concern. He explained that the system was built around a particular vision of typical progress, leaving little room for variation. When students fell outside those expectations, they were quickly marked as problems, regardless of cultural context, trauma history, or disability related differences. As Carlos stated,

When normal is defined in one way, difference automatically becomes deficit. Anyone who does not match that definition is immediately seen as behind or problematic. There is no space in the system to recognize different ways of learning or developing. Instead of asking whether the expectations are too narrow, the system assumes the student is the issue. Over time, that framing shapes how decisions are made and how students are understood.

Postsecondary participants observed that these dynamics did not end with K to 12 schooling but extended into transition and accommodation contexts. Lisa described how AI driven tools frequently framed disability through a lens of risk rather than access, noting that students were often flagged based on what they might not be able to do. These flags shaped advising and monitoring practices in ways that limited opportunity, shifting the focus from removing barriers to managing perceived liabilities.

Michael echoed these concerns, describing how predictive analytics reinforced narrow models of persistence and productivity. Systems measured students against an idealized version of success, positioning those who deviated as at risk. Disability, in this context, appeared not as a call to provide support but as a warning sign embedded within the data. As Michael noted, the system “measures who fits the model,” leaving little space to question whose model that is.

Collectively, these accounts demonstrate how artificial intelligence systems reinscribe ableist and racialized norms by transforming historical inequities into data driven truths. Rather than challenging exclusionary practices, AI often magnified them by embedding normative assumptions into automated processes. This theme illustrates how data systems function as conduits through which longstanding hierarchies of race and ability are reproduced within special education governance, shaping who is monitored, who is trusted, and who is positioned as educable.

Theme 4: Policy Silence and Unclear Accountability

Uncertainty about responsibility surfaced repeatedly as participants described how artificial intelligence shaped special education decision making. Rather than pointing to formal policy guidance, participants spoke about moments when questions emerged after decisions had already been made, often in response to concerns raised by families, auditors, or students themselves. In these moments, artificial intelligence was powerful enough to influence outcomes but elusive enough to evade clear accountability. Participants consistently described a landscape in which AI shaped practice without being clearly owned, governed, or regulated.

Paula characterized this ambiguity as one of the most troubling aspects of AI integration. When decisions informed by data systems were questioned, she explained, responsibility rarely rested in a single place.

There is no clear answer about who owns the decision. It feels like responsibility is shared until something goes wrong, and then it becomes very unclear who is accountable. The district points to the system, the system points back to how it was used, and individuals are left in the middle. No one claims ownership when the outcome harms a student. In the end, the people closest to the student carry the weight of decisions they did not design.

Whether accountability belonged to the district, the vendor, or individual professionals remained unclear. As Paula put it,

It always feels like it lands back on the people closest to the student, even though we did not design the system. We are expected to explain and defend decisions that were shaped by tools we had little say in choosing. When families push back, it is our judgment that is questioned, not the system itself.

Jane similarly described how the absence of policy guidance left professionals navigating ethical and legal dilemmas without institutional support. She explained that AI tools were frequently adopted without clear direction about how their outputs should be interpreted, contextualized, or challenged.

We are told to use the system, but we are not told what to do when the system does not align with what we see. There is no guidance for how much weight we are supposed to give our professional judgment in those moments. It creates hesitation because going against the data feels risky without policy backing. You end up trying to balance what you know is right for the student with what the system expects you to do, Jane stated.

This lack of clarity, she noted, became particularly fraught when decisions were later scrutinized by families or oversight bodies. In those moments, professionals were expected to justify outcomes without being able to point to clear policy guidance about how algorithmic information should have been weighed. Questions about fairness, appropriateness, and process were redirected toward individual decision makers rather than toward the systems that shaped those decisions. As a result, accountability became personalized rather than institutionalized, leaving professionals exposed while the role of artificial intelligence remained largely unquestioned.

At the school level, Carlos described how accountability gaps shaped everyday practice. Although AI tools strongly influenced evaluation and referral processes, there were few formal mechanisms for questioning their assumptions or limitations.

If the system flags a student and something goes wrong, it is not clear who is responsible. The data is treated as authoritative, but no one takes ownership of its impact. When outcomes are questioned, the responsibility shifts to the individual who followed the recommendation. That uncertainty makes it harder to challenge the system, even when concerns are present, he explained.

Without policy protection, challenging algorithmic outputs felt risky for many participants. Carlos explained that questioning the system required taking on personal and professional vulnerability without institutional backing. “It is easier to follow the data than to challenge it without support,” he noted. Over time, this dynamic reinforced compliance with algorithmic recommendations even when they conflicted with professional judgment.

Mary echoed these concerns, emphasizing that policy silence often resulted in heightened individual liability. She described a contradiction in which educators were expected to rely on AI outputs while remaining personally accountable for outcomes.

We are told the system is a guide, but when there is a problem, it is not the system that gets questioned. The responsibility falls on the individual who followed the recommendation. There is very little space to point back to the tool or its design. That makes you think twice about raising concerns, even when something does not feel right.

This dynamic created pressure to align professional decisions with algorithmic recommendations, even when those recommendations raised concerns about equity or appropriateness. Participants described weighing their professional judgment against the perceived authority of the system, often choosing compliance to avoid risk. Over time, algorithmic outputs came to function as the safest option, regardless of whether they fully reflected student needs. This pressure narrowed the space for advocacy, particularly for students whose experiences did not fit standardized models. As a result, equity concerns were often subordinated to the demands of consistency, documentation, and institutional protection.

Postsecondary participants observed similar patterns as students transitioned beyond K to 12 contexts. Lisa explained that AI supported accommodation systems often operated with minimal transparency, leaving uncertainty about how decisions were made and how they could be contested. “Students assume the system is the policy,” she noted, “but when they appeal, there is no clear place to point to.” Michael further described how accountability was frequently redirected toward students themselves. “When students struggle, the focus is on their persistence,” he explained, “not on whether the system set them up to fail.”

These accounts reveal how policy silence allows artificial intelligence to exercise authority without bearing responsibility. In the absence of explicit governance structures, accountability is displaced onto individual professionals and students rather than embedded within institutional or policy frameworks. This theme highlights how the lack of clear policy guidance enables AI systems to shape special education decision making while obscuring where power resides and who is responsible when harm occurs.

Discussion

Rather than demonstrating longitudinal accumulation, the findings point to strong cross sector resonance in how artificial intelligence shapes special education practice across P to 16 education. Drawing on interviews with six participants and analysis of institutional documents, including district guidance, procurement materials, and internal protocols, the findings show how AI functions as a policy actor in practice. Across roles, participants described AI as embedded within data systems and decision-making processes rather than as a clearly defined policy initiative. District leaders emphasized efficiency, compliance, and risk management, a framing reinforced in procurement and guidance documents that prioritized data integration and accountability while offering limited attention to disability equity or due process.

At the school and postsecondary levels, professionals described how AI shaped evaluation, placement, and accommodation processes, often privileging algorithmically generated indicators over contextual judgment. Internal protocols positioned these indicators as default reference points, constraining professional discretion despite formal commitments to individualized decision making. Across interviews and documents, AI systems were treated as technical enhancements rather than policy interventions, contributing to ambiguity around accountability and equity. Analysis revealed four interrelated themes: AI adoption driven by efficiency and compliance, the elevation of algorithmic authority, the reproduction of ableist and racialized norms, and the absence of clear governance structures. Together, these findings demonstrate how artificial intelligence reshapes decision making and power within special education systems.

Policy Implications and Recommendations

The findings of this study demonstrate that artificial intelligence is already functioning as a policy actor in special education systems, despite the absence of clear regulatory frameworks that address equity, accountability, and civil rights. Without intentional policy intervention, AI risks reinforcing ableist and racialized exclusion under the guise of efficiency and neutrality. The following recommendations outline concrete actions at the federal, state, and institutional levels to ensure that artificial intelligence governance aligns with the foundational principles of special education and disability rights.

Federal Policy Actions

At the federal level, artificial intelligence governance must extend beyond alignment with existing civil rights statutes to include explicit regulatory mechanisms addressing the risks of algorithmically mediated decision making. The U.S. Department of Education should both clarify how AI tools intersect with the Individuals with Disabilities Education Act, Section 504, and the Americans with Disabilities Act and establish binding standards requiring pre implementation approval of AI systems used in eligibility, placement, services, and accommodations, including documentation of use, decision logic, and equity risks. Federal policy should mandate Algorithmic Impact Assessments prior to deployment, evaluating potential disparate impacts using quantitative and qualitative data, with submission tied to federal funding and ongoing updates. Concurrently, procedural safeguards should ensure rights to explanation, contestation, and human review, while independent third-party audits should assess bias, system design, and civil rights compliance, with findings publicly reported to strengthen transparency and accountability.

State Policy Actions

At the state level, departments of education should translate federal standards into enforceable regulatory frameworks governing the procurement, implementation, and use of artificial intelligence in special education. States should require formal approval processes prior to adoption, including vendor documentation, data governance plans, and evidence of alignment with equity goals, with review by multidisciplinary panels with expertise in special education, data ethics, and civil rights. Policies must also protect professional discretion by prohibiting AI as the sole determinant in decision making, requiring documentation when judgment diverges from algorithmic outputs, and shielding educators from liability when acting in the interest of equity. In addition, states should establish clear data governance and transparency requirements, limit predictive analytics in high stakes decisions, and ensure families have accessible information and meaningful participation. Finally, it is critical that accountability systems incorporate equity focused indicators tied to AI use, shifting emphasis from compliance to substantive outcomes.

Institutional Actions in K to 12 and Postsecondary Settings

At the institutional level, districts and postsecondary institutions must take proactive steps to govern AI use responsibly. Schools and universities should establish interdisciplinary AI governance committees that include special education professionals, disability services staff, legal experts, families, and individuals with disabilities. These committees should review AI tools prior to adoption, monitor their impact, and create clear protocols for addressing concerns. Institutions should also develop written policies that clarify accountability when AI informs decision making. These policies must specify who is responsible for outcomes, how decisions can be challenged, and what recourse is available to students and families.

Clear accountability structures are essential to preventing the displacement of responsibility onto individual professionals or students, a pattern repeatedly identified in the findings. Professional development is another critical institutional responsibility. Educators, administrators, and disability services staff must be trained not only in how to use AI tools but in how to critically evaluate them. Training should address algorithmic bias, disability justice, and the limits of data driven decision making, equipping professionals to resist compliance-oriented pressures that undermine equity.

Implications for Future Research

The findings of this study highlight the need for longitudinal research that examines how artificial intelligence shapes special education outcomes over time. Rather than focusing solely on adoption or short-term effects, future studies should trace how AI informed decisions related to eligibility, placement, surveillance, and accommodations accumulate across educational transitions. Such research would illuminate the long-term consequences of AI as a policy actor within special education systems and clarify its role in shaping access, persistence, and postsecondary outcomes for students with disabilities.

Future research must also center the voices of students with disabilities and their families in analyses of artificial intelligence and educational policy. Participants’ accounts suggest that AI reshapes how decisions are communicated, contested, and experienced, yet these impacts remain understudied. Grounded in DisCrit’s emphasis on lived experience and counter storytelling, research that foregrounds student and family perspectives can challenge dominant narratives of neutrality and efficiency, particularly for racially marginalized communities historically subjected to surveillance and exclusion.

There is also a critical need to expand DisCrit informed policy research within the field of educational technology. Much existing scholarship remains technocratic, prioritizing innovation and efficiency while underexamining power and inequality. Applying DisCrit more broadly would shift attention toward the policy conditions that enable educational technologies to reproduce ableist and racialized inequities. Comparative and cross sector studies could further clarify how differing governance structures shape the equity implications of AI in special education.

Limitations of the Study

This study has several limitations that should be considered when interpreting the findings. The sample consists of six participants drawn from two states in the southeastern United States, representing primarily suburban and urban contexts, which limits geographic diversity and does not capture rural perspectives or broader national variation. While the study provides in depth insight across key professional roles, the small sample size is not intended to support generalizability but rather to offer interpretive and context specific understanding. Additionally, the data reflect participant perspectives and institutional documents within particular settings, which may not fully represent how artificial intelligence is enacted across all special education systems. The cross sector design includes both K to 12 and postsecondary contexts; however, the study does not employ a longitudinal approach and therefore cannot empirically trace cumulative impacts over time. Finally, as artificial intelligence tools and policies continue to evolve, the findings represent a snapshot of practice during a period of rapid technological adoption and may shift as governance structures develop.

Conclusion

This study demonstrates that artificial intelligence is already reshaping special education policy and practice in ways that extend far beyond technical innovation. Through a Disability Critical Race Theory lens, the findings reveal how AI functions have the potential to reinforce ableist and racialized bias and marginalize professional judgment. Within primary and postsecondary institutions, AI driven systems influence eligibility, placement, surveillance, and accommodation decisions, often compounding inequities rather than mitigating them. Without explicit equity centered governance, these technologies risk undermining the civil rights foundations of special education by obscuring responsibility and normalizing exclusion under claims of neutrality and efficiency. This study calls for urgent policy attention that recognizes artificial intelligence as a matter of educational justice, demanding regulation, accountability, and research frameworks that center disability, race, and lived experience.