Generative artificial intelligence (AI) is rapidly reshaping work, education, and civic life, but its benefits remain unevenly distributed where access and infrastructure are limited. Mississippi offers an important case for examining this problem. Federal Communications Commission (2024) data show that several counties, many in the Black Belt or Delta region, have fixed broadband coverage rates below 10%, with many others near 20%. Combined with one of the nation’s highest poverty rates (Benson, 2024), these conditions constrain residents’ ability to adopt emerging technologies and participate in digitally mediated learning and work. In higher education, Skinner, Burtch, and Levy (2024, p. 837) report that Mississippi ranks last nationally in in-home broadband access among undergraduates, falling 16.1 percentage points below the national median. These barriers restrict remote collaboration, digital learning, and the development of AI-relevant skills.
Digital inequity in Mississippi reflects not only infrastructure gaps, but also affordability constraints, uneven device access, and varied levels of digital literacy. The C Spire Rural Broadband Consortium (2020) reported that broadband availability across the state averaged 72% in 2020, yet actual usage was only 17.8%, with some rural counties reporting usage rates as low as 3.6%. Research on digital inequality consistently shows that such conditions reduce educational opportunity, contribute to lower academic performance and well-being, and limit students’ ability to participate fully in digitally mediated learning environments (Hampton et al., 2023, pp. 6–10, 31–36). In contexts where connectivity is unstable and technology access is uneven, AI learning initiatives designed for well-resourced environments may fail to translate into sustained participation or practical adoption. In this study, AI access is understood not only as access to devices, connectivity, and digital tools, but also as access to meaningful opportunities to learn, apply, and critically engage AI in ways that are relevant to everyday life and work.
At the same time, Mississippi is home to a network of Black educational institutions, community organizations, and STEM initiatives working to broaden participation despite structural constraints. Historically Black Colleges and Universities (HBCUs), K–12 educators, nonprofit organizations, and related STEM pathways have developed locally grounded strategies for supporting students and communities under conditions of underinvestment (Kirtman et al., 2021; Lee et al., 2019). In this context, questions of AI access are closely tied to longer histories of educational inequality, uneven digital infrastructure, and community-based innovation in Black educational and civic institutions (Federal Communications Commission, 2024; Skinner et al., 2024).
Prior research highlights the importance of early exposure, mentorship, and locally relevant instructional design in strengthening STEM identity and persistence among students of color (Lee et al., 2019; Perez et al., 2014; Zhou et al., 2017). Emerging K–12 AI education models similarly emphasize creativity, ethical reflection, and constructionist learning as core elements of early AI engagement (Ali et al., 2019, p. 2), while national initiatives such as AI4ALL (2022) reflect growing momentum toward equity-centered AI education. Together, these strands of research suggest that equitable AI participation requires more than technical content alone; it also depends on instructional designs that support belonging, relevance, and meaningful use.
Formally founded in June 2023, the Mississippi AI Collaborative (MAIC) addresses the intersection of technological change and structural inequity through a community-based approach to AI literacy and workforce development (data.org, 2025). MAIC expands access to generative AI training, builds educator capacity, and supports pathways into technology careers through mentorship and project-based learning. In contrast to institution-centered models or technically intensive bootcamps, MAIC emphasizes local expertise, cultural relevance, and place-based design. By centering Mississippi’s racial, geographic, and infrastructural realities, MAIC offers a model for how AI learning can be organized around the lived conditions that shape participation in historically under-resourced communities.
Systematic reviews of AI literacy education indicate that the field remains dominated by school- and university-based interventions focused on conceptual models, competency frameworks, and structured curricular modules (e.g., Casal-Otero et al., 2023; Kong et al., 2021). These interventions overwhelmingly take place in well-resourced educational settings with stable digital infrastructure. Across these reviews, there is limited empirical evidence on AI literacy initiatives situated in rural regions, Black-majority communities, or contexts shaped by persistent digital inequity. Moreover, much of the AI ethics education literature remains centered on classroom-based case studies, simulations, and curricular modules, with comparatively less attention to community-responsive engagement with the sociotechnical histories and structural constraints that shape learners’ experiences (Biagini, 2025; DiPaola et al., 2024; UNESCO, 2022). As a result, there remains a need for research that examines how AI literacy is taken up in settings where access barriers, local context, and culturally grounded learning conditions are central to program design.
This study examines MAIC as an instrumental case that offers insight into how equitable AI education can be designed for and with marginalized communities. The analysis is guided by three research questions: (1) How do participants describe changes in their confidence, understanding, and use of AI tools? (2) How do culturally grounded and community-responsive teaching strategies shape engagement and learning? and (3) Which elements of MAIC’s design are perceived as most effective in supporting AI learning and workforce readiness?
Literature Review
Research on AI literacy has grown rapidly in recent years, producing a diverse and increasingly complex body of work that spans K–12 education, community engagement, informal learning, and digital ethics. Systematic reviews reinforce that there is no single agreed-upon definition of AI literacy; instead, the field reflects “a rich mosaic of interpretations that mimic the evolving nature of AI itself” (Biagini, 2025, p. 20). Biagini’s (2025) systematic review, which screened 323 publications and identified 87 relevant studies, maps the wide definitional range of AI literacy across technical, ethical, and socio-cultural domains.
As scholars have noted, AI literacy involves not only foundational knowledge of AI systems but also critical awareness of their ethical and societal implications, reflecting broader work on responsible and critical AI use (Druga et al., 2019, 2022; Wong et al., 2020). Definitions across the field suggest that AI literacy encompasses the knowledge required to understand AI, the operative skills needed to engage it, and the awareness necessary to critically evaluate its uses and consequences. Several conceptual strands shape this growing field. Some frameworks emphasize technical competencies such as computational thinking, machine learning concepts, neural networks, and coding (Ng et al., 2021b; Su & Zhong, 2022). Others adopt a broader view that foregrounds ethics, culture, and socio-technical understanding (Cuomo et al., 2022; Hermann, 2022; Yi, 2021). Long and Magerko’s (2020) competency model further underscores that AI literacy must prepare learners not only to interact with current tools but also to navigate evolving technological environments over time.
This scholarship suggests that AI literacy is inherently interdisciplinary, requiring both practical fluency and critical discernment. At the same time, these definitions also imply that AI literacy is not developed in the abstract. Learners must have opportunities to encounter AI tools, experiment with them, and reflect on their uses in context. In this sense, AI literacy depends not only on curriculum, but also on the material and institutional conditions that make sustained participation in AI learning possible.
Scholars mapping the international landscape of AI-in-education research identify a persistent structural issue: publication patterns remain heavily concentrated in a small set of countries. Garzón, Patiño, and Marulanda note that this “uneven geographic representation raises concerns about whether AI tools can effectively meet the diverse needs of educational systems worldwide” (Garzón et al., 2025, p. 9). Similar trends appear in the AI literacy literature, where China and the United States account for much of the empirical research, while Africa and Latin America remain underrepresented. Uneven representation may also operate within national contexts, where research is more likely to emerge from well-resourced institutions and major metropolitan regions than from smaller, historically under-resourced states such as Mississippi. This matters because the conditions under which AI literacy develops are shaped by geography, infrastructure, institutional support, and local educational ecosystems. Where broadband access, device availability, and instructional capacity are uneven, learners do not enter AI education on equal terms, and models developed in more resourced environments may not translate easily into other contexts.
Several strands of research also emphasize the importance of early, developmentally appropriate exposure to artificial intelligence concepts while pointing to persistent limitations in how AI learning is designed and assessed. In their systematic review of K–12 AI education, Casal-Otero et al. (2023) observe that although many learning experiences focus on technical or conceptual content, “there were hardly any experiences that assessed whether students understood AI concepts after the learning experience,” underscoring a persistent lack of evaluative rigor in the field (p. 1). The authors further note that instructional designs remain fragmented and inconsistent, calling for competency frameworks that are “modular, personalized, and adjusted to the conditions of the schools” (p. 11), particularly where educational resources and technological infrastructure vary widely. This emphasis on variation in school conditions is especially important for understanding the relationship between digital access and AI literacy. AI learning does not occur under uniform conditions; it is shaped by whether learners have the time, connectivity, devices, and instructional supports needed to move from exposure to meaningful practice.
A growing line of work links AI literacy with civic and ethical education. DiPaola et al.'s (2024) case study of the AI and Human Rights curriculum shows that middle school students can develop nuanced understandings of privacy, discrimination, and safety when AI literacy is taught through project-based civics lessons and legislative simulation. Their findings illustrate the value of interdisciplinary approaches that treat AI as a sociotechnical system rather than a purely technical domain. Biagini’s (2025) systematic review similarly distinguishes between technical AI literacy frameworks rooted in computational thinking and operational skills and holistic frameworks that foreground ethical, social, and cultural dimensions. As Biagini notes, holistic models better prepare learners “to adapt and thrive amidst societal transformations” (p. 22). Others argue that curriculum design efforts must balance both orientations, consistent with recommendations by Kong et al. (2021), Laupichler et al. (2022; Laupichler, Hadizadeh, et al., 2022), and UNESCO (2022). These studies underscore that meaningful AI literacy education requires integrating technical competence with civic and ethical reasoning, but they also suggest that such integration depends on learning environments that are responsive to the realities learners bring with them.
Curriculum development challenges therefore persist globally. Although many curricula foreground technical content, fewer provide practical strategies for teaching AI ethics despite broad consensus on its importance. This gap supports calls for interdisciplinary AI education that embeds ethical reasoning and sociotechnical analysis directly within course design. UNESCO’s (2022) international guidelines reinforce this concern by advocating for curricula that integrate digital literacy, computational thinking, data literacy, and ethical reasoning to prepare learners to navigate AI’s societal impacts. Hsu et al. (2022) similarly argue that AI efforts must be responsive to “local concerns,” noting that researchers often need to “fine-tune existing models or build new pipelines to fit local needs” (p. 5).
These studies suggest that AI literacy cannot be understood only as a matter of curriculum or content. In historically under-resourced settings, digital access is not merely a background condition for AI literacy; it shapes whether learners can encounter AI tools, participate in sustained and flexible learning opportunities, and develop the practical and critical capacities that AI literacy requires. This relationship is especially important in places such as Mississippi, where uneven broadband access, resource disparities, and longstanding educational inequities may influence not only whether AI tools are available, but also whether learners can engage the kinds of community-based, contextually meaningful experiences through which AI literacy develops.
Conceptual Framework
Because MAIC was developed through practitioner expertise, community partnerships, and iterative evaluation rather than a predetermined theory of change, this study draws on three frameworks as a composite interpretive lens for understanding how participants described AI learning, engagement, and perceived impact: cognitive apprenticeship, culturally relevant pedagogy (CRP), and Afrofuturist educational theory. These frameworks are not treated as a priori design assumptions; instead, they function as analytic scaffolds for interpreting participants’ accounts of how learning was supported, why engagement felt accessible and meaningful in context, and how participants oriented toward AI as a tool for creativity, agency, and community futures.
Cognitive apprenticeship conceptualizes learning as the development of complex cognitive skills through guided participation in authentic tasks, with expert thinking made visible through modeling, coaching, scaffolding, articulation, reflection, and exploration (Collins et al., 1991, p. 6). This lens is especially relevant for AI literacy, where learners must develop practical fluency (e.g., prompting, iterative refinement, output evaluation, and transfer of AI use into workplace or instructional routines) alongside conceptual understanding. In MAIC, it supports interpretation of participant accounts emphasizing iterative practice, feedback cycles, and real-world application, including mentor modeling, coached troubleshooting, scaffolded progression toward independent use, and reflective learning strategies.
CRP complements this emphasis on supported skill development by foregrounding the conditions under which learning becomes meaningful and sustaining. CRP argues that effective learning environments support academic success while affirming cultural identity and cultivating critical consciousness about inequity (Ladson-Billings, 1995). In STEM and AI contexts, where “neutral” norms can reproduce exclusion through deficit narratives and misalignment with learners’ lived realities, CRP highlights how belonging and legitimacy are produced through relational practices and ideological shifts rather than surface-level inclusion. Ouedraogo-Thomas and Miles (2026) demonstrate how anti-Blackness operates in STEM spaces through biased practices and deficit framings that restrict belonging and opportunity, while Anderson and Deil-Amen (2024) show that culturally responsive mentoring can humanize STEM pathways for low-income community college students but remains insufficient without culturally responsive curriculum and instruction. In this study, CRP informs interpretation of participant descriptions of approachable learning environments, place-based relevance tied to Mississippi communities and professional roles, and relational learning practices that reduced barriers to entry.
Afrofuturist educational theory further extends the composite lens by foregrounding futurity, imagination, and technological agency as dimensions of equitable STEM learning. McGee and White (2021) argue that Afrofuturism enables learners to envision Black presence in their visions of the STEM future, aligning with participants’ emphasis on AI as a tool for creativity, authorship, and community futures. Eseonu and Okoye (2024) similarly position Afrofuturism as a critical qualitative inquiry orientation for liberation, offering conceptual tools for interpreting how Black communities envision alternative futures amid structural inequality. In MAIC programming, this lens supports interpretation of participant narratives that framed AI not only as a productivity tool but also as a medium for creativity, experimentation, and community benefit, particularly in a state where infrastructural and racialized inequities can constrain technological participation.
The composite lens conceptualizes equitable AI literacy as simultaneously (a) skills-based, developed through guided practice and feedback; (b) culturally sustaining, shaped by place-based relevance, belonging, and relational learning conditions; and (c) future-facing, oriented toward imagination, agency, and community benefit. Applied to MAIC as an instrumental case, this framing supports interpretation of how community-rooted AI education can expand technical participation while also strengthening belonging and agency within emerging technological futures. Consistent with the study’s inductive thematic approach, these frameworks were applied after initial coding to deepen analytic interpretation rather than to predetermine categories.
Methods
Design
This study uses an instrumental case study design to explore how a community-based AI education initiative can illuminate broader issues of equity, access, and culturally grounded learning. The Mississippi AI Collaborative (MAIC) was selected because it operates in a historically under-resourced, predominantly Black region of the U.S. South where broadband inequities, economic disparities, and uneven STEM opportunities shape how learners encounter emerging technologies. MAIC’s emphasis on mentorship, local relevance, and flexible access conditions made it a productive site for examining how community-rooted AI education unfolds in a racially and infrastructurally constrained environment.
Research Orientation
This study is grounded in a constructivist-interpretivist orientation, which views knowledge as co-constructed through experience and reflection. Several team members were directly involved in MAIC’s design or facilitation, offering contextual insight but requiring reflexivity. Participant responses were treated as situated narratives that reflected how learners made meaning of their experiences. The goal of this study was interpretive understanding rather than validation or causal inference.
Participants
The broader population served by MAIC includes K-12 educators, college students, higher education faculty, adult learners, youth, and small business owners across Mississippi, many in communities where broadband access and computer science opportunities remain limited. Program participation records indicate that MAIC’s Year One activities engaged 871 AI-thon learners (75% women; 68% Black/African American), over 1,400 educators through the Educator Accelerator and Fellowship pathways (60% ages 35–54), and 86 AI Agency clients (71% women; 95% Black/African American). MAIC’s workforce-aligned pathways also trained 30 college students identifying as Black/African American.
Although MAIC reached over 3,000 educators, students, and community members statewide during Year One, this study draws on a targeted analytic sample of participants who completed follow-up evaluation activities. The analytic sample includes 154 individuals who submitted survey responses and/or written reflections during MAIC’s first year: 123 K–12 and workforce educators participating in the Educator Accelerator and Fellowship programs, 21 youth and adult learners enrolled in AI-thon course pathways, and 10 participants in the AI Agency and Catalyst programs (five college apprentices and five business or nonprofit partners providing structured feedback; Appendix A). To protect participant privacy, demographic details are reported only in aggregate at the program level, and no subgroup-level demographic breakdowns are reported where small counts could increase re-identification risk.
Data Collection
Three primary data sources informed the study: (a) post-program surveys (5-point Likert-scale items and open-ended prompts), (b) reflective journals from college apprentices, and (c) program artifacts such as curricular materials and facilitator notes. Surveys were distributed immediately after program sessions, reflective journals were collected at the mid-point and completion of applied project work, and artifacts were drawn from MAIC’s planning and instructional archives. Not all participants contributed each data type due to program-specific evaluation instruments; indicator prevalence estimates therefore reflect combined evidence available for each participant across surveys and written reflections. All evaluation activities followed informed consent procedures, and responses were anonymized prior to analysis.
Across instruments, data collection focused on participants’ self-reported learning experiences, perceived usefulness of AI tools, and contextual barriers to participation. Likert-style items were measured on a 5-point scale (1-5; 5 = strongly agree) and assessed outcomes such as confidence applying AI knowledge, perceived relevance to work or teaching, and intention to continue using AI tools. Open-ended prompts asked participants to describe key takeaways, intended applications, and anticipated challenges. Apprentice journals documented tasks completed, skills practiced, feedback received, and shifts in understanding over time. Program artifacts included instructional slide decks, course pathway materials, facilitation notes, example prompts, and program planning documents that contextualized MAIC learning activities and delivery.
Analysis, Trustworthiness, and Ethics
Qualitative analysis followed a two-phase, codebook-informed process that combined structured keyword flagging with contextual manual coding, consistent with hybrid approaches to qualitative thematic analysis that integrate a priori categories with iterative interpretive refinement (Fereday & Muir-Cochrane, 2006; Appendix B). First, the research team developed initial theme areas and keyword indicators based on program goals and recurring concepts identified in a shared subset of responses (e.g., confidence, hesitancy, mentorship, practical application, community relevance, ethical reflection, and access barriers). Responses containing these indicators were then flagged for review, but keyword filtering was used only to prioritize cases for closer examination and did not determine final coding decisions. Next, two researchers reviewed flagged responses in full context to confirm thematic fit, refine code definitions, and assign theme labels, with responses allowed to contribute to multiple themes. Final prevalence estimates therefore reflect interpretive manual coding rather than keyword matches alone.
Because evaluation instruments varied by program, indicator prevalence was computed using a unified evidence rule. A participant was counted as meeting an indicator if (a) they endorsed a conceptually aligned Likert item at 4-5 (agree/strongly agree) when that item was present in their instrument and/or (b) their open-ended response or journal reflection contained clear, contextually supported language consistent with the indicator definition (Appendix B). Keyword filtering was used only to flag responses for review and did not determine coding. Two researchers independently reviewed and coded flagged responses; differences were resolved through discussion and codebook refinement. All prevalence values are descriptive for the analytic sample and should not be interpreted as causal effects.
To enhance trustworthiness, the team iteratively refined code definitions through constant comparison and consensus discussions. Triangulation across surveys, apprentice journals, and program artifacts strengthened interpretive credibility by aligning participant accounts with documented program structures. Because several researchers were involved in MAIC design and facilitation, reflexivity was supported by prioritizing participant language and collaboratively reviewing theme boundaries and counterexamples.
All evaluation activities followed informed consent procedures. Demographics are reported in aggregate to minimize re-identification risk, and quotes use pseudonyms with broad role or program identifiers. The project was conducted as program evaluation for continuous improvement.
Program Description
As outlined in the introduction, Mississippi’s AI landscape is shaped by racialized broadband inequities, uneven computer science opportunities, and persistent economic constraints. At the same time, HBCUs, community organizations, and grassroots technology efforts have built a foundation for expanding access under these conditions (Kirtman et al., 2021; Robison, 2018). The MAIC was developed within this context to expand access to AI literacy and workforce development.
MAIC’s approach is grounded in the state’s racial, infrastructural, and geographic realities. Its programming integrates four interconnected components: community AI literacy, educator capacity building, workforce-aligned apprenticeships, and long-term innovation infrastructure. Community AI literacy is introduced through the AI-Thon, a mobile-friendly, asynchronous set of courses that provide foundational AI concepts and applied tasks for K-12 learners, educators, entrepreneurs, and adult learners. In-person learning events, including workshops hosted by The Bean Path, a local nonprofit, and the Southern Spark Conference, extend these efforts through hands-on exploration and community engagement.
Educator capacity building is supported through the Educator Accelerator and Research Fellowship programs, which blend in-person workshops, virtual sessions, and classroom-focused “AI Labs.” Led by the Mississippi Computer Science Teachers Association and Jackson State University, these initiatives emphasize manageable integration strategies for teachers who may be new to AI tools and include opportunities for fellows to develop lesson plans and lead peer training within their districts.
Workforce-aligned training and apprenticeships are offered through the Mississippi Coding Academies and the AI Agency and Catalyst pilot programs at Jackson State University. College apprentices collaborate with Mississippi businesses, nonprofits, and educational institutions to design AI-supported tools and workflows, including prompt engineering, chatbot development, automation pipelines, and generative storytelling. In the Catalyst model, apprentices are linked to partner organizations to support exploratory AI projects. Mentorship, reflection, and iterative feedback are embedded across both programs.
Finally, MAIC invests in long-term innovation infrastructure. The initiative has developed an open-access digital platform for AI courses, begun exploratory efforts to train language models using Mississippi-based data, and launched a practitioner-research journal dedicated to inclusive AI education. These efforts reflect a commitment not only to increasing AI literacy, but also to fostering local authorship, critical engagement, and sustained technological participation. By situating AI education within Mississippi’s specific sociotechnical landscape and embedding programs in existing community networks, MAIC offers a contextually grounded model for expanding equitable participation in emerging technologies.
Findings
Analysis of surveys, apprentice journals, and program materials surfaced four interconnected themes describing how participants experienced MAIC programming; descriptive participant-level indicator prevalence is reported for the analytic sample (N = 154).
Note. Percentages reflect the share of participants in the analytic sample (N = 154) whose survey responses and/or written reflections provided evidence consistent with each indicator definition (manual interpretive coding; see Appendix B). Participants were counted as meeting an indicator if evidence appeared in either Likert endorsements (ratings of 4–5 on a 5-point scale) and/or qualitative reflections. Values are reported descriptively and should not be interpreted as causal effects or population estimates. The AI-thon flexibility statistic (63%) is calculated within the AI-thon subgroup only (n = 21).
Building Confidence
Across participant groups, the most consistent pattern was increased confidence using AI tools. Across the analytic sample, 87% of participants met the indicator for increased confidence after completing MAIC programming. Participants frequently described a shift from unfamiliarity to routine use and often referenced specific tools such as ChatGPT, Microsoft Copilot, Gemini, Magic School, and Brisk. In addition, 76% described concrete practical use cases, and 42% reported sharing AI knowledge with peers (Appendix B).
Participants often linked growing confidence to practical application in their instructional and professional roles. Denise (Educator Accelerator Participant; K-12 teacher), for example, described integrating Microsoft Copilot into both instructional planning and professional research tasks: “I’ve been using what I learned about Co-Pilot to help me brainstorm class activities. It has improved my classroom environment. I’ve also used it to help me do some preliminary research and brainstorming on research topics.” This suggests that confidence was reinforced when AI could be applied to familiar, immediate workplace tasks.
Participants also framed confidence gains as increased clarity about AI’s capabilities and limits. Tanya (Educator Accelerator; secondary teacher) described the program as strengthening conceptual understanding while sharpening practical judgment: “I like that it covered a broad range of AI uses/forms. I had used machine learning before and had a broad understanding of how AI trains, but this training really solidified that for me and in so doing made it clear what AI can and cannot be used for.” Other responses emphasized confidence as a psychological shift from apprehension to ease. Dr. Lewis (higher education participant), for instance, reflected: “I was intimidated at first. Now it’s second nature.” These reflections indicate that confidence involved both greater conceptual clarity and increased ease of use.
Confidence was also connected to workload reduction and instructional efficiency. Andrew (Educator Accelerator; elementary teacher) described using AI to “help me craft ideas, edit, and save time,” positioning AI as a practical support for daily responsibilities. Similarly, Ms. Reed (Educator Accelerator; instructional coach/specialist) emphasized that MAIC’s accessible instruction “helped to expel any fears that new users have or may have.” These comments further suggest that confidence was closely tied to participants’ sense that AI could make everyday work more manageable.
One educator reflection illustrates how AI confidence gains were linked to sustained professional demands. As Angela (Educator Fellow; experienced K-12 educator) explained:
For three years, I researched and studied teacher burnout… This understanding created a huge passion to learn all things AI to help combat the unnecessary stress that extra workloads cause.
College apprentices similarly described confidence emerging through applied, real-world learning. Tony (AI Agency apprentice) stated: “This experience has made me job-ready. I feel confident using these skills in real projects.” These qualitative findings indicate that confidence gains were not limited to tool awareness, but were grounded in participants’ ability to apply AI to authentic tasks in their instructional and professional roles through guided practice and real-world use.
Relevance, Identity, and Culturally Grounded Learning
Participants consistently emphasized the importance of instruction that felt approachable, locally meaningful, and connected to the realities of Mississippi schools and workplaces. In the analytic sample, 29% of participants explicitly emphasized community relevance or local grounding (Table 1). Many participants also described MAIC’s instructional tone as accessible for learners with limited prior exposure to AI tools. Early-career educator Kayla (Educator Accelerator), for example, described the training as well-calibrated to her entry point: “It was straightforward, engaging, and very approachable. Not too much. Just right for where I’m starting.” This suggests that participants valued learning environments that matched their entry points and reduced barriers to engagement.
A related pattern was that participants valued instruction that reduced intimidation while emphasizing practical strategies. Alana (Educator Accelerator workshop attendee; middle school educator) described the workshop as beginner-friendly and confidence-building: “The workshop did a great job of explaining AI in a beginner friendly way and helped to expel any fears that new users have or may have. You have to be intentional with the way that you word things to get the results that you want.” This reinforces the importance of approachable instruction that translated AI into usable practice.
Participants also described relevance in terms of immediate classroom applicability. Nicole (Educator Accelerator; elementary school librarian) highlighted the usefulness of tool exposure for differentiated instruction:
This was an excellent presentation that introduced us to the AI platforms available and ways in which we could use them to generate lesson plans, activities, differentiated lessons and much more. The presenter was amazing at explaining how to use all of the tools and gave suggestions for ways to use AI at different levels.
These comments suggest that participants experienced relevance not only through the content itself, but through learning environments that made AI feel accessible and applicable within their own professional contexts.
Although ethics and equity concerns were referenced less frequently (9%; Table 1), those who raised them did so in ways that linked AI learning to broader community impact. Dr. Carter (higher education faculty participant), for example, reflected: “What stuck with me most was thinking about who builds AI and how it affects our communities. That should be part of every class.” Similarly, Monique (Educator Accelerator; K–12 educator) emphasized the importance of “keeping in mind the known concerns surrounding AI in education, especially the equity issues, bias & misinformation.” These responses suggest that, for some participants, AI literacy was meaningful not only as a technical skill set but also as a way of engaging broader ethical and community concerns.
For youth participants, AI’s creative potential also supported engagement. Jayden (AI-thon learner; youth participant) summarized this takeaway as “something we can use to make stuff.” Collectively, these reflections suggest that MAIC supported engagement by presenting AI as both usable and meaningful, connected to participants’ instructional roles, community needs, ethical considerations, and, for some learners, creative possibility.
Mentorship, Apprenticeship, and Learning Through Doing
Mentorship and guided practice were central to how participants described learning through MAIC programming. Within the analytic sample, 41% referenced mentorship or guided support (Table 1). This theme was particularly prominent in workforce-aligned pathways, including MAIC’s AI Agency initiatives, where apprentices described learning through iterative practice, feedback, and partner-facing tasks.
Nia (AI Agency apprentice; college student) described mentorship as enabling understanding and skill development: “The mentor I had was incredibly patient and helped me connect the dots.” Darius (AI Agency apprentice; college student) similarly emphasized the importance of iterative feedback cycles: “The feedback cycles and mentorship really helped me grow. I’ve learned more from working with this than in any class.” These comments highlight the importance of guided support in helping participants build skills through practice and revision.
Mentorship also supported participants’ emerging sense of themselves as capable technology users and knowledge sharers. Tony (AI Agency apprentice; college student) reflected: “I never thought I’d use AI like this. Now I’m showing others how.” Partner organizations echoed the reciprocal value of this applied learning model. Ms. Harper (AI Agency client; small business owner) valued “the eagerness of the apprentice wanting to help my small business… and providing valuable resources I can refer to in the future.” These accounts suggest that mentorship strengthened both technical confidence and participants’ sense of their ability to support others.
A longer partner reflection illustrates how mentorship-supported implementation translated into concrete outcomes:
As a small business in Jackson, before joining the AI Agency, I would spend weeks trying to update the website on my own. With the Agency’s guidance, the process became quicker and clearer, and for the first time, I felt my online presence was actually reaching new customers. (Mr. Jones, AI Agency client; small business owner)
Educators participating in MAIC’s professional learning pathways similarly described structured support as enabling pedagogical transfer. As Latonya (Educator Fellow; K-12 teacher) wrote: “Being part of the fellowship gave me structure and confidence. It helped me not just learn AI but learn how to teach it.” These findings indicate that mentorship operated as both a learning scaffold and a mechanism for translating AI knowledge into real-world instructional and community-facing applications.
Structural Barriers and Uneven Access
Despite MAIC’s accessibility-oriented design, structural constraints shaped how participants experienced AI learning. Approximately 22% of participants referenced technical barriers such as broadband limitations, device constraints, or connectivity challenges (Table 1). Dr. Ellis (K-12 educator; rural-serving context), for example, cautioned: “Without continued broadband access and funding, maintaining these innovations will be challenging.” This highlights the continued importance of infrastructural support even within accessibility-oriented programs.
Flexibility emerged as a key support for participation. Across the analytic sample, 35% highlighted flexibility or mobile-first access as important (Table 1). Among surveyed AI-thon learners (n = 21), 63% reported that mobile or flexible access was necessary due to schedules, limited connectivity, or shared device use. Sabrina (AI-thon learner; adult participant) wrote: “It was easy to do at my own pace and on my phone.” Andre (AI-thon learner; adult participant) similarly emphasized: “I enjoyed all the training. It was easy to understand, and I could do it at my own pace.” These comments suggest that flexibility functioned as a practical condition of participation rather than simply a preferred delivery format.
Participants also framed flexible delivery as supporting engagement under varied participation constraints. Grace (Educator Accelerator; director/educator; early childhood/child care setting) noted: “I appreciate the virtual learning opportunity, since I’m not doing large crowds at this time due to underlying health concerns. I learned a lot and hopefully will have the chance to participate in future workshops.” This indicates that flexible design also supported participation under personal and logistical constraints.
Finally, participants connected AI learning to the practical realities of workload and time. Denise (Educator Accelerator; K–12 teacher) summarized this framing succinctly: “AI is a great resource that we can add to our teacher toolbox.” Participants also emphasized that sustained adoption requires continued infrastructure investment. As Marcus (AI-thon learner; adult/workforce participant) stated: “AI is not going anywhere… we need to learn how to use it as a tool.” These findings suggest that MAIC’s flexible delivery model supported participation under structural constraint, while participants simultaneously recognized that long-term AI engagement depends on broader digital inclusion conditions.
Discussion
The findings of this case study illuminate how community-embedded AI education can create meaningful shifts in confidence, skill development, and critical engagement among learners in structurally under-resourced contexts. In Mississippi, where broadband inequities, uneven access to computer science education, and persistent economic disparities shape technological participation, MAIC offers one instructive model for how community-driven design can expand pathways into AI literacy. The discussion below interprets these findings through the lenses of cognitive apprenticeship, CRP, and Afrofuturist educational theory, while also examining how the results align with and extend existing research on AI education and digital inequality. These frameworks are used here as interpretive lenses for understanding participant accounts rather than as claims about a predetermined program design model.
Participants’ substantial increases in AI confidence suggest that exposure and structured support can counteract the effects of systemic exclusion. Many participants described beginning with hesitancy or fear and gradually developing comfort and competence. This shift reflects the demystification process central to cognitive apprenticeship, where modeling, guided practice, and feedback scaffold learners toward independent application (Collins et al., 1991; Minshew et al., 2021). The MAIC programs functioned as structured apprenticeships in which participants learned by doing, supported by mentors who made invisible processes visible. Interpreted through this lens, increased confidence is not only an attitudinal outcome, but evidence of supported skill acquisition through coached participation, which is particularly important in rural Mississippi, where technical fluency carries broader implications for community empowerment and workforce mobility.
The centrality of relevance and cultural grounding further reinforces the importance of contextualized learning in technology education. Participants consistently emphasized that MAIC’s instructional examples felt accessible, locally meaningful, and designed for their realities rather than for abstract or idealized learners. These patterns are consistent with CRP, which emphasizes learning environments that affirm learners’ identities, build competence, and cultivate critical awareness. MAIC’s place-based examples and community-facing projects supported these aims by situating AI not as a distant concept, but as a tool for addressing everyday challenges in Mississippi schools, small businesses, and community organizations. In this sense, CRP helps interpret participants’ engagement as shaped not only by content and delivery, but by whether the learning environment communicated that AI knowledge is legitimate and attainable within Black and rural communities. For Black educators and youth, especially those navigating constrained digital environments, the sense that “AI is for us” is itself transformative and a counter-narrative to deficit framings that have historically marginalized their participation in STEM (Ouedraogo-Thomas & Miles, 2026, p. 2).
Participants’ reflections on ethics, bias, and community impact reveal another layer of culturally grounded engagement. While only a subset explicitly articulated these concerns, those who did connected AI learning to questions about who designs technology and how it affects marginalized communities. These reflections resonate with scholarship critiquing algorithmic bias and calling for intersectional, justice-oriented approaches to AI development (Cardenas & Vallejo-Cardenas, 2019; Vethman et al., 2025). They also reflect emergent ethical reasoning fostered through contextualized learning environments where learners have space to consider AI’s implications for their own communities. In Mississippi, such ethical engagement is particularly salient given the broader context of digital inequity and the risk that emerging technologies may reproduce existing structural harms.
Future-oriented imagination also emerged as a meaningful throughline in participants’ experiences. As the findings suggest, apprentices and youth described AI as a tool for creativity, experimentation, and self-expression, and some positioned themselves as emerging experts who could introduce AI tools to peers or community members. These responses are consistent with Afrofuturist educational theory, which frames Black learners as agents capable of imagining and shaping technological futures (McGee & White, 2021; Williams, 2024). Afrofuturism offers a way to understand how participants experienced a sense of authorship and possibility through AI learning, a notable outcome given Mississippi’s longstanding structural barriers to STEM participation. Rather than encountering AI as a fixed system produced elsewhere, learners described orientations toward AI as something they could adapt, remix, and use in service of community goals; an interpretive pattern consistent with Afrofuturist approaches to futurity, liberation, and technological agency (Eseonu & Okoye, 2024).
At the same time, the findings highlight persistent systemic limitations. Participants repeatedly referenced broadband issues, device shortages, and access challenges that shaped the conditions of their learning. These constraints reflect broader patterns of infrastructural inequality documented in Mississippi and similar regions (C Spire Rural Broadband Consortium, 2020). They also confirm that even the most responsive educational programs cannot compensate for systemic deficits in digital infrastructure without coordinated public investment. However, MAIC’s flexible, mobile-compatible design demonstrates how programs can adapt to these constraints to reduce barriers to participation. The program’s responsiveness underscores the importance of structurally aware design rather than one-size-fits-all models that assume stable access and uniform technological readiness.
Overall, the findings illustrate a model of AI education that is relational, contextualized, and future oriented. Recent analyses warn that generative AI may widen racial wealth and opportunity gaps without deliberate investment in inclusive strategies (McKinsey & Company, 2023). The insights from this study contribute to broader efforts to design equitable AI education models in underserved regions by showing how mentorship-driven learning (cognitive apprenticeship), culturally sustaining engagement (CRP), and future-oriented technological agency (Afrofuturist educational theory) can operate together as complementary dimensions of community-based AI literacy. Findings from this instrumental case suggest design considerations for comparable under-resourced contexts and the broader implementation of AI education. MAIC’s emphasis on local partnerships, mentorship structures, culturally tailored curricula, and equitable access to digital infrastructure demonstrates a promising pathway for inclusive STEM engagement. These strategies can serve as a model for regions with similar structural barriers, where traditional, top-down approaches may fall short.
Limitations
MAIC’s evaluation instruments were designed for program improvement rather than research, resulting in varied survey and journal prompts that limit direct comparison across educators, apprentices, and community learners and narrow the claims that can be made about group differences. Participation was also self-selected, meaning respondents may have been more motivated or more positively inclined toward AI learning than nonrespondents. The study relies on self-reported reflections rather than observational or performance-based evidence, so long-term classroom use, workplace impacts, and sustained skill application remain unknown.
Future Directions
AI education in under-resourced regions requires models that respond directly to local context, community expertise, and the structural conditions shaping technological access. MAIC’s outcomes show that relying on stable broadband, advanced devices, or prior computer science exposure risks reinforcing existing inequities, while flexible, mobile-accessible delivery and community-rooted mentorship can create meaningful entry points for learners historically excluded from technological innovation. Participants’ reflections on bias, representation, and community impact illustrate that AI literacy must extend beyond technical skill-building to include culturally grounded ethical inquiry, aligning with calls for intersectional and justice-oriented AI education (Vethman et al., 2025). Mentorship emerged as central to developing confidence and practical fluency, echoing research on cognitive apprenticeship and STEM identity formation. Future work should examine long-term impacts on classroom practice, workforce mobility, and continued STEM engagement; compare outcomes across regions; and consider how broadband expansion, device access, and technical support shape participation. Additional innovation in Afrofuturist, narrative-based, and community-authored AI projects may further strengthen learners’ sense of agency. As generative AI reshapes institutions and labor markets, the key question becomes not only whether communities can access AI education, but whether such models equip them to shape more equitable technological futures.
Conclusion
The MAIC illustrates how AI education can be reimagined through community partnership, cultural relevance, and embedded mentorship. In a state where digital inequity shapes everyday access to technology, MAIC demonstrates that meaningful engagement with generative AI is possible when learning environments are designed around the realities, strengths, and aspirations of local communities. Participants across programs reported greater confidence, clearer understanding of AI tools, and deeper ethical awareness, outcomes that were reinforced by hands-on practice, contextually grounded instruction, and opportunities to see themselves reflected in technological futures.
Interpreting these findings through the lenses of cognitive apprenticeship, CRP, and Afrofuturist educational theory reveals how participants developed both practical skills and a sense of agency. Mentorship and guided practice supported authentic skill development; culturally grounded examples made AI feel relevant and attainable; and opportunities for creativity helped learners imagine themselves as contributors to their communities’ digital futures. These insights emphasize that equitable AI education is not merely a matter of content delivery but of relational, contextual, and imaginative design.
Although MAIC cannot resolve Mississippi’s structural barriers alone, it offers a feasible model for advancing AI literacy in under-resourced contexts. Its layered and collaborative approach, combining flexible formats, community-rooted pedagogies, and locally meaningful applications, charts a path for educators, policymakers, and technologists seeking to create inclusive and culturally sustaining AI ecosystems. As generative AI continues to evolve, the imperative is clear: invest in programs that center people, culture, and place so that historically marginalized communities are not only included in technological change, but equipped to shape it on more equitable terms.