Higher education struggles to keep pace with trends in technology, which has often trailed the industry significantly (Maslej et al., 2023). ChatGPT reached 100 million users in a mere matter of two months, which is exponentially less time than past services reached the same number of users, like Instagram and Facebook (Heath, 2023; Porter, 2023). Employers today want college graduates with experience and skills in GenAI tools (Cengage Group, 2023). Yet, the slow tendency for higher education to adapt to new technologies means a growing inequity is emerging between those who can use the tools effectively and those who cannot.

AI and equity have become a pressing concern, highlighted by organizations like the Bill and Melinda Gates Foundation, which considers AI equity among its top priorities. The foundation emphasizes not only equitable access to AI technologies but also the importance of co-creating tools that foster fairness and inclusion (Suzman, 2023). While the biased output of GenAI models is often cited as being an issue (D’Agostino, 2023), the issues of educational equity are much larger. While the last two years show an increase in research to understand the potential impact of AI technologies, particularly in the sphere of higher education (Petricini et al., 2023; Wu et al., 2024; Zipf et al., 2024), still missing from this growing body of literature are answers to the questions about equity (Bond et al., 2024).

The study of equity in the use of GenAI within higher education cannot be disentangled from the broader context of existing inequities that pervade society, higher education, and technology. These inequities manifest in multiple layers, influencing who has access to AI tools, who benefits from their usage, and whose voices are amplified or marginalized through these technologies. Central to these inequities is the concept of information inequity, which highlights the uneven distribution of access, skills, and agency in using AI tools, often reflecting, and reinforcing broader societal and educational disparities. The introductions of many past technologies have often led to both advocates and resistors, including the support and disdain for tools like the blackboard, calculators, and even the Internet, partly because the power dynamic between the instructor and student changed (see Davidson, 2022). As AI changes the power dynamics of knowledge, higher education structures will undergo fundamental and deeply transformative shifts that will lead to the ‘privileging of certain forms of knowledge and communication’ (Kurban & Şahin, 2024, p. 34) . Higher education has the potential to reduce or increase inequities of GenAI use for all faculty, students, and staff.

Artificial intelligence continues to grow in use and acceptance, making this a timely investigation for how key groups in the higher educational milieu perceive the impact AI has on equity. Our guiding research question was: What perceptions do students, faculty, and staff hold on the impact AI has on educational equity?

Important to this study is the definition of ‘GenAI usage’. We define this term to mean interacting with tools to refine questions, seek answers or solutions, assist in brainstorming or idea generation, or any other action within a designated tool.

The site location for this study was a large, research-intensive, historically white institution (HWI), located in the mid-Atlantic region of the United States. Selected items are reported in this study as part of a larger research project. After approval from the institutional review board, we sent emails to unit leaders across the institution to share with faculty, students, and staff in their areas. The emails included a short introduction, a link to the Qualtrics survey tool, and implied consent and minimal risk with participation information.

Participants answered the open-ended question, ‘How does generative AI impact educational equity?’ Responses were analyzed with a combination of deductive and inductive thematic coding, in which the latent level was used to underscore and connect deeper meaning to the responses (Braun & Clarke, 2006). The codebook was developed a priori based on the five dimensions of information equity by Lievrouw and Farb (2003) (Table 1). We considered Access to mean the ability to pay for, use, or reach the GenAI technology and concepts related to socio-economic status and ability to obtain digital technology and devices. Content meant pertaining to the output, data, or deliverable from a GenAI tool, whereas Context was where GenAI tools were located within a learning environment and higher education, implying change to society, higher education, teaching and learning, and social systems. Skills related to the teachable abilities or knowledge about the use of GenAI tools and knowing how to perform actions and tasks. We also used Skills when people described using GenAI in different ways, either beneficial or harmful. Lastly, Values implied the principles or standards that govern personal and societal belief and behavior; determining if something is good or bad, beneficial, or harmful, right, or wrong, and highlights the ethics and morality of using the GenAI tool. By using these dimensions, we were able to maintain an equity lens on the responses.

Our data suggested additional dimensions not represented in Lievrouw and Farb (2003). For those responses unrelated to equity, we used N/A. Additionally, our data showed people were unsure of how equity matters or plays a role with GenAI tools and applied the code Uncertainty in those instances.

We utilized a thematic analysis approach to the data. Thematic analysis is a flexible analysis for any type of qualitative data that is not bound by a certain theory (Braun & Clarke, 2006). Data was loaded into an Excel spreadsheet and shared with the team. The research team discussed the five dimensions, and two researchers coded independently during round one, resulting in an unacceptable level of inter-rater reliability (κ = .57). The two researchers discussed the codes, agreeing on the meaning and extension of the dimensions, and re-coded the open-ended responses with an acceptable level of inter-rater reliability of (κ = .89).

Limitations of this study exist. Data that was coded by our method may not yield the same results if a different lens is applied. For example, some data were not returned as codable because the response had nothing to do with educational equity but rather served as a commentary on teaching and learning in general terms. While still very valuable to the larger discussion around GenAI, these points were not included in our analysis. Comparison to institutions with different faculty, student, and staff populations should be made with care. Our site location provides paid versions of GenAI tools to the entire community, which not all institutions can afford. Likewise, institutions with

different policies, either prohibiting or integrating GenAI tools in the classroom, institutional strategy, or operations, might find other concerns related to equity left unknown by our sample.

Our results show that perceptions of equity related to GenAI in higher education are shaped by multiple factors of influence, extending beyond traditional concerns like biased content and physical access. While access remains an important issue, it is clear a more nuanced understanding of equity is needed—one that considers the broader context of skills, support systems, and institutional policies. Our findings highlight that equity concerns are not limited to who can afford or use GenAI tools but also include how these tools are applied and the role of literacy development in navigating them. Importantly, values and ethics are emerging as critical yet understudied dimensions of AI and equity. The diverse perspectives of students, faculty, and staff suggest that a singular focus on access and bias may obscure other essential factors that influence how GenAI tools shape educational equity.

While our student sample size was small (n=31) and not representative of the student population, the student responses were surprisingly focused on values, not access like the faculty and staff members. This finding was surprising, given the presumption of cheating frequency with GenAI tools (McMurtie, 2024; Wiley, 2024). This finding should not go unnoticed, as it appears integrity is important to them when using these tools; however, a discrepancy of what ‘cheating’ means may exist between students and faculty, adding to the confusion of what should be considered as academic dishonesty (Dahl & Waltzer, 2024). Eaton (2024) calls for moving past the traditional practices associated with academic integrity by decolonising what is considered acceptable academic output. By moving past these traditions, society will need to reevaluate what is ‘ethical’ and what is ‘integrity’ as our tools become more embedded in our future (Eaton, 2023). More research should be done with larger samples to see if and how ‘values’ remain a key aspect of students’ GenAI use and students’ definitions of cheating and academic dishonesty.

The dimension of access is crucial in revealing disparities in who can afford and effectively use GenAI tools. While faculty and staff tended to focus on physical access and financial barriers, Van Dijk (2017) critiques this view as narrow and argues the digital divide is often misunderstood as a simple gap between two groups. He explains that ‘the metaphor suggests a simple distinction between two divided groups with a yawning gap between them,’ which ignores the complexities and relative nature of digital inequalities (p. 1). Societal systems of oppression and privilege have been evident with digital access for decades as the importance of technology-literacy gained attention (Gorski, 2003). The early digital divide was rooted in physical access but neglected the deeper, more nuanced issues related to skills, social support, and usage (Gorski, 2003; Van Dijk, 2017).

In our research, we see that this neglect persists in the rhetoric associated with AI equity. Van Dijk’s notion of ‘skills access’ is particularly relevant here, referring to the ‘digital and media literacy’ skills necessary to use technology effectively (p. 7). Students who lack these skills are disadvantaged, even if they have physical access to the tools and will not resolve the divide (Gorski, 2003). A more multidimensional view of access, one that encompasses both the operational skills to command digital media and the substantive skills required to engage meaningfully with it, is needed. Additionally, Van Dijk highlights ‘usage access,’ where differences in how frequently and creatively individuals use technology further exacerbate inequities. Faculty who discourages students from exploring advanced uses of GenAI tools may inadvertently deepen socio-cultural inequities. Those with higher education tend to use digital tools for ‘capital-enhancing goals’ like career and academic growth (Van Dijk, 2017, p. 8), while others may only engage with simpler applications, such as entertainment (Bonfadelli, 2002; Cho et al., 2003; van Deursen & van Dijk, 2014; Van Dijk, 2017; Van Dyjk, 2005; Zillien & Hargittai, 2009).

As Van Dijk notes, barriers to access often go ‘beyond’ physical access to include ‘computer anxiety’ and ‘technophobia,’ which are particularly prevalent among underrepresented groups such as seniors, people with low education levels, and certain female populations (2017, p. 5). In our research, faculty, and staff express fears about biased GenAI content, but these concerns may overlook other critical barriers to equity, like the psychological and social obstacles that inhibit full engagement with these tools. The focus on content bias, though important, may obscure the more pressing need for skill-building and equitable support structures that empower all students to use and learn these tools meaningfully. When these skills go undeveloped, people are unable to gather and use information for informed decision-making, a form of information poverty (Marcella & Chowdhury, 2020), and an outcome of information inequity

In conclusion, while traditional concerns like physical access and biased content remain relevant, our findings reveal a deeper need to consider how factors such as skills, institutional support, and values influence equitable access and the use of AI tools. Equity is not merely about whether students can afford or physically access GenAI but also about the extent to which they possess the necessary skills, motivation, and opportunities to engage meaningfully with these technologies. To address the AI equity gap identified in this study, we propose that future research should prioritize the development of a multidimensional framework that integrates key equity dimensions such as access, skills, values, and institutional support. This framework should extend beyond traditional binary notions of access to encompass nuanced aspects like skills access, usage access, and literacy development as outlined by (Van Dijk 2017). Additionally, it must consider the role of institutional policies in fostering an inclusive culture where diverse values and ethics are acknowledged and integrated into AI usage norms. Institutions should take a comprehensive approach that integrates AI literacy with an awareness of equity issues, fostering environments that actively support all members in overcoming barriers, whether they are technical, social, or psychological. A critical component of this framework would involve co-designing interventions with stakeholders— students, faculty, and staff—to ensure diverse perspectives shape strategies to bridge inequities. Such a framework could include mechanisms for fostering AI literacy through equitable support systems, embedding ethical AI usage in institutional curricula, and creating tools that mitigate psychological and social barriers like technophobia. By broadening the focus of equity discussions to include these dimensions, research would provide a robust foundation for developing policies and practices that promote meaningful engagement with GenAI across diverse educational contexts. Future research should expand on our findings, examining diverse institutional contexts and larger populations to deepen our understanding of how GenAI shapes educational equity and how we can address the potential gap.