From Prisons to Programming: Fostering Self-Efficacy via Virtual Web Design Curricula in Prisons and Jails

2.1 Digital Technology in Correctional facilities

Hawley et al. [35] highlight that a significant proportion of incarcerated people have low education levels, exhibit high drop-out rates, lack lifelong learning competencies, and exhibit low motivation. As Barros et al. [7] report, prison learning environments must respect different types of learning and paces, while encouraging participation. However, while incarcerated people are a heterogeneous group [19], their diverse learning needs are not typically accounted for in prison environments [38].

Analyzing the requirements for internet-enabled learning in prison contexts, previous studies have outlined the need for courses to promote skills development, interactivity, self-confidence, and motivation [55], echoing findings for adult learners in general [54]. Identifying teaching practices to foster these effects, a study in UK prisons by Gray et al. [34] reports that active classroom participation is key to fostering students’ sense of self-determination and confidence. Other work on digitization in prisons suggests a link between digital contact with the outside world and greater self-esteem [45], and the capacity for internet-enabled audiovisual communication to grant access to programming that increases post-release circumstances, facilitating rehabilitation and reducing recidivism [52].

Reporting in 2018, Reisdorf and Rikard [69] discuss the scarcity of US computer literacy programs inside correctional facilities, and state that none cover how to navigate the internet or use modern technologies. The ubiquitous digitization of society over the past 20 years has led researchers to suggest that denying access to these tools while incarcerated effectively constitutes a second punishment, resulting in incapacitation and further social exclusion post-release [89]. This concern is compounded following the United Nations declaring internet access a human right [43]. Now, governments have begun a shift in policy. In the UK, groups such as Justice Digital, working with the Ministry of Justice, and The Center for Justice, a domestic non-profit, are exploring ways to drive digital change in UK prisons [30]. In the United States, the California Department of Corrections and Rehabilitation announced in 2022 a forthcoming roll-out of 30,000 secure laptops to incarcerated students with access to pre-approved internet sites [59].

With digitization set to expand in carceral environments, there is high need to study implementations that meet key metrics linked to positive post-release outcomes, such as self-efficacy. However, many studies investigating the need for digitization in correctional facilities are review-based, observational, or theoretical in nature, without directly evaluating the efficacy of new interventional measures inside prisons. Two key reasons for this to date are digital security and federal regulation. First, access to computers and internet in U.S. correctional facilities have been tightly controlled, and second, incarcerated people’s classification as a federally-protected population of research participants creates practical hurdles to research in this area [60]. While a number of studies assess the benefits of digitization in general, few studies assess the efficacy of digital-based computer programs in particular. In this paper, we describe a new digital literacy program, conducted virtually in five correctional facilities, and use surveys to suggest what specific course designs may contribute to success in terms of student self-efficacy.

2.2 Mitigating incarceration in HCI

HCI researchers have previously cited racial disparities in prisons to argue that "all HCI research must be attuned to issues of race; participation of underrepresented minorities must be sought in all of our activities" [61]. Despite this, Verbaan et al. [78] report that prison contexts are under-explored from an HCI perspective; only a handful of publications actually perform work within that context, and the incorporation of technology remains in its infancy in terms of incarcerated people’s access. As discussed, primary reasons for this are digital security and federal regulation. However, research addressing issues on incarceration, a topic that has become central to discussions around justice in the United States, has in the last few years been increasingly explored in HCI research.

Ogbonnaya-Ogburu et al. [62] found that returning citizens who experienced long prison sentences emerge with almost no digital literacy skills, and piloted a six-week digital literacy course in which they found that returning citizens were hungry for digital literacy skills and differed in their needs in key ways from other marginalized groups. Other researchers examined participatory design processes of a virtual reality (VR) reentry training program with a women’s prison to create narratives that reflect the group’s particular challenges [75]. However, the scope of this work did not include an evaluation of its efficacy. Anuyah et al. [3] introduced a framework based on Maslow’s hierarchy of needs to conceptualize how lack of digital literacy and technology access impose barriers on incarcerated and other marginalized peoples. Interviews conducted with 75 women transitioning out of prison showed how low self-efficacy and digital literacy are major issues faced by women-in-transition in particular [73]. More generally, digital literacy has been identified as a key barrier to accessing social services for formerly incarcerated people and other vulnerable community members [4]. In our work, we report on a web programming course that we designed and taught in 5 correctional facilities, and whose outcomes on self-efficacy and digital literacy we have analyzed, providing a vital link between incarcerated people’s experiences and HCI practitioners.

Other HCI researchers have addressed incarceration from a different approach. While our work addresses interventions that may help people stay out of prisons, other researchers have worked on mitigating bias in sending people into them. HCI researchers have explored the perceptions of algorithmic unfairness in traditionally marginalized communities, for whom consequences can include arrest [85]. Street-level Algorithms [1] investigated judicial bail–recommendation systems, interrogating biases that included disproportionately recommending jail for people of color. Veale et al. [77] interviewed machine learning pracitioners to gauge how designers can imbue values like fairness and accountability into algorithmically-informed public decisions on issues including sentencing and resource allocation within prisons. Bauman et al. [9] developed machine learning-driven intervention models to connect social and mental health workers with at-risk populations in need of care, in order to avoid incarceration. Rather than mitigating algorithmic bias in arrests and sentencing, our contribution complements the existing body of HCI work by revealing interventions that foster successful reentry once sentences are served.

2.3 The Case for Self-Efficacy and Digital Literacy

Prison-based education programs are effective for improving reentry outcomes. In a 2013 meta-analysis, researchers found that incarcerated adults who participated in reading and/or mathematics courses had a 13% lower rate of recidivism compared to those who did not [20]. A more recent meta-analysis spanning research published from 1980-2017, found an even more significant result: incarcerated students in correctional education programs were 32% less likely to recidivate [11]. In this same analysis, researchers found that incarcerated people who participated in correctional education had 12% higher odds of obtaining post release employment compared to those who did not. However, it is unclear whether the association between prison-based education programs and these positive outcomes is a causal one.

It may be the case that the positive link between correctional education and reentry outcomes is driven by psychosocial characteristics or cognitive skills, as is suggested by research in traditional post-secondary academic settings. Self-efficacy, or the belief in one’s ability to execute behaviors necessary to meet a goal [6], and digital literacy, the competencies needed to perform tasks in digital environments, are two possible mediating factors that have been associated with improved academic performance. A systematic review synthesizing research from 2003 to 2015 showed that self-efficacy was moderately positively correlated with academic performance [37] and this relationship extends to asynchronous, online courses [87]. Similar effects have been reported for digital literacy and performance [88], as well as for digital literacy and students’ intentions to pursue further education [53]. Overall, the benefit of high self-efficacy and digital literacy skills for academic achievement is well-supported by research in non-carceral academic settings.

A limited number of studies have investigated psychosocial characteristics or cognitive skills of incarcerated populations due to the traditional emphasis on outcomes like recidivism. Those studies have found a positive relationship between postsecondary correctional education and self-efficacy, cognitive skills, self-evaluations, and self-esteem. More focused research indicates that self-efficacy is a key mediator of the effect of correctional education on both proximal outcomes like academic success and more distal outcomes like recidivism [2, 42]. Reentry programs with a focus on digital literacy acquisition have led to significant reductions in recidivism [84]. However, with research aiming to establish relationships between education and outcomes, a key factor that distinguishes studies with incarcerated people is the challenge of setting up research programs and acquiring large sample sizes. To name one example, a recent meta-analysis [11] found no significant effect of computer-assisted instruction on academic performance (compared to traditional instruction), however the dataset was too limited to draw conclusions about this relationship. Our study is an important contribution to the sparse literature investigating both psychosocial factors and digital literacy skills in the context of incarceration.

3.1 Structure and curriculum

The program runs for 12 weeks, meeting twice weekly for class, and once weekly for office hours. Each class is 2 hours long, consisting of a 30 minute lecture followed by 90 minutes spent in breakout rooms with a Teaching Assistant (TA). Each breakout room was staffed by 1 Teaching Assistant, working with 2-4 students to implement and debug code. TAs were recruited predominantly from our institution’s computer science department, and included PhD students, faculty, staff, and alumni. We also invited formerly incarcerated graduates of our program to TA, whose input helped align our program’s teaching to the current students’ needs. The choice of using breakout rooms, combining low student-teacher ratios with project-based coursework, is modelled on studio-based instruction. Within the HCI community, researchers have studied how to optimally teach computer science, with studies demonstrating that students overwhelmingly endorse studio-based design to learning involving project-based open-ended design problems [24]. The studio course’s attendant teaching mode of mentor and mentee can be found wherever design is taught [24]. For courses on subjects like web programming, it is this design orientation that makes HCI-related courses exceedingly suitable for Studio-based instruction [40]. To foster Studio-like mentor-mentee relationships, we combined our project-based curriculum with dedicated recruitment drives to achieve low student-teacher ratios. Our 2022 class had 16 Teaching Assistants and 45 students enrolled at course start, for an average breakout room student-teacher ratio of 2.8:1. In 2023, the program had 21 Teaching Assistants, with 57 students enrolled at course start, for an average breakout room student-teacher ratio of 2.7:1. In correctional facilities in particular, studies reveal the importance of "meeting the students where they were" in terms of academic abilities, and of providing detailed and constructive feedback [46]. Our program’s template for recruiting large numbers of outside Teaching Assistants may also mediate broader benefits; as Link et al. [46] report, one way to diminish the “us versus them" attitudes exhibited by non-incarcerated people is to allow these individuals to engage in shared learning experiences that create a sense of community. In summary, our student-teacher ratios were designed to facilitate individualized teaching that catered to the breadth of student backgrounds and fostered an inclusive learning environment. The curriculum and individualized teaching was designed to support learners with no previous computer experience and required no academic prerequisites.

Figure 2:

View Figure

The first 6 weeks of the 12-week curriculum centers on the fundamentals of HTML, CSS and Javascript, including branching statements, loops, and functions. The latter 6 weeks center on a capstone project. For this, students apply their learnings to build websites addressing a pressing social issue of their own choosing to affect impact in their communities. Popular issues chosen by students included domestic violence, gun control, and addiction—issues which many students have first-hand experiences with. Their websites centered around offering support, outlining factual information, and raising awareness about their topic, using their technical skills to provide a vital service while reconnecting with the communities to which they will eventually return. The homepages from six websites programmed by students in our 2023 cohort are shown in Figure 2. As a college-accredited program, all passing students can utilize their credits to enroll in one of the degree-granting partner universities post-release.

3.2 Planning and implementation

To foster transparency and encourage further research in correctional education, we report on some of the logistical, security, and administrative challenges with bringing educational programming to facilities, and with collecting data on these programs. These challenges include enabling computer instruction, internet access, and bringing men and women into a virtual room to learn together. While data for these topics is sparse, and the landscape is rapidly changing, we summarize key challenges and place them in context.

4.1 Ethical Considerations

We followed the ethical guidelines put forth by the Perspectives on Ethical Inquiry [67] developed by The Inside-Out Center’s Evaluation and Research Committee—a leader in evaluations for prison-based educational programming—to ensure our research exceeded the federal regulations for Respect for Persons, Beneficence, and Justice. Our procedures also ensured that the program instructors were not the researchers, to help mitigate any concerns of coercion or other negative perceptions, and the course instructors were never aware of the students’ participation status. These procedures were communicated to all participants before the study was carried out. We received approval of our research design from our institution’s Institutional Review Board (IRB). Incarcerated people are considered a vulnerable population for research purposes, but participation in this survey posed only minimal harm because their participation in the education program was voluntary, participation in the survey was voluntary, the questions did not ask about sensitive topics, and their names were kept confidential. All students were asked to participate if they chose to provide responses, and no responses were excluded from the study. Participants were not compensated for participation in the survey, and the instructors observed secure data handling procedures.

4.2 Facility and Participant Selection

We carried out our studies with all 5 participating correctional facilities, which we refer to here alphabetically as Facilities A through E. Facility A included participants from both men’s and women’s units, including residents across low, medium and high security custody categories. Facility B is a medium security facility, and included only women. Facility C is a minimum security facility, and also included only women. Facility D is a medium security facility, and included participants from both men’s and women’s units. Finally, Facility E is a mixed medium and minimum security facility, and included only men. Specific student enrollment in the class was controlled by the correctional facilities’ education staff, who filter enrollment based on conflicts in students’ schedules and recent misconduct; participation in our study was offered to all enrolled students. The studies were carried out as virtual surveys with consenting participants. All participant data is aggregated and anonymized, and we received approval of our research design from our institution’s Institutional Review Board (IRB).

We selected the 5 participating correctional facilities in an effort to reflect the representation in the general prison population. In particular, the participating facilities covered a broad range of security classifications, including a jail, a reentry center, as well as minimum and medium security prisons. This selection was also informed by the values expressed by HCI researchers that "participation of underrepresented minorities must be sought in all of our activities" [61]. With our original partners exhibiting under-representative populations of Black students (at 2% and 9.5%, respectively [70]), we expanded our partnership in 2022 to reach a more racially representative group of students. Our 2023 study had 29.4% of respondents identifying as Black, 8.7% as multiracial, with 8.8% preferring not to say. This is approximately reflective of the US general prison population, which is 34% Black according to a 2014 study [57]. Design Justice [16], a practice aimed at ensuring "a more equitable distribution of design’s benefits and burdens" and "meaningful participation in design decisions", also influenced our selection of participating correctional facilities. Because women constitute a relatively small portion of the U.S. prison population, women are at higher risk of being discounted in correctional research aimed to foster positive post-release outcomes. To reach these students, our program partnered with women’s correctional facilities. In our 2022 and 2023 programs, the percentages of initially enrolled students who identified as women were 52.4% and 49.1%, respectively. Drawing further from Design Justice practices, our program also included student alumni from previous years who returned to serve as TAs, contributing to the program design and content.

4.3 Participant Demographics

4.4 Structure and protocol

5.1 Thematic Analysis of Participant Feedback

This section offers a thematic analysis of the qualitative data gathered from post-course surveys completed by 34 incarcerated students in our 12-week, college-accredited web design course from 2022. We focus on understanding the impact of HCI elements on self-efficacy and learning experiences within the unique carceral context.

5.2 Statistical Analysis

Table 2 shows the sample sizes (n), means (M), and standard deviations (SD) for student outcomes, for both pre-test and post-test. Note that differences in sample size reflect incomplete data. The mean scores measuring general self-efficacy and the CPSES subscales all increased between pre- and post-test, but the relatively large standard deviations indicate substantial variation in scores, perhaps due to our small sample size (n=28-36).

Table 3 shows the results for one-sided paired t-tests to determine whether student outcomes at the end of the term were improved relative to the start of the term. For this test, any participant who did not complete both pre- and post-tests for an item were excluded from the analysis. This left 15 individuals who completed a pre-test and a post-test for measures of general self-efficacy, logic CPSES, and algorithms CPSES, and 14 individuals who completed the CPSES subscales for cooperation, control, and debugging. We applied an alpha level of 0.05 for all t-tests. The mean difference shows in this case that all CPSES subscales increased between pre- and post-test, while the general self-efficacy score fell marginally by 0.01, but exhibiting the largest variability in our data. Contrary to our hypothesis, neither general self-efficacy nor any CPSES subscales were statistically significantly different at post-test compared to pre-test. The difference in means test found that there was no statistically significant difference between the mean scores of self-efficacy and CPSES subscales, except for self-efficacy in algorithms (p value = 0.01). We cannot reject the null hypothesis to claim that the program increased students’ self-efficacy, but the small sample size provided limited statistical power to make such a determination. Replication studies can incrementally add to the research on this subject and provide a basis for meta-analyses with increased statistical power.

Qualitative data from the 2022 course also suggested that the more experience participants have, the less their self-efficacy would improve in subsequent courses. We measured previous experience in two ways: the number of previous programming courses and the number of previous courses that used computer applications. Although we planned to conduct a one-way analysis of variance to determine if changes in self-efficacy may depend on prior experience, we did not have sufficient data to conduct these analyses.