TL;DR
This paper evaluates the "Caregivers and Machine Learning" (C&ML) program, an AI education initiative designed for caregivers on parental leave. The study found that the program successfully increased AI literacy and confidence among participants, with a notable percentage transitioning into AI-related roles or further education, highlighting the effectiveness of inclusive, targeted educational interventions.
What Problem Does This Paper Address?
The paper addresses the significant underrepresentation of caregivers, particularly women, in AI-related fields. Despite AI's growing importance across industries, systemic and individual barriers prevent many from entering or advancing in these domains. This gap is exacerbated by the demanding nature of caregiving responsibilities, which can lead to career interruptions and a perceived "leaky pipeline" effect. The C&ML program was developed to tackle this by offering an accessible and supportive AI education specifically tailored to the needs of caregivers on parental leave, aiming to equip them with the necessary skills and confidence to re-enter or advance in the tech workforce.
Methods
The study employed a mixed-methods approach to evaluate the C&ML program. Data collection involved two stages:
- Post-Program Electronic Survey: A 41-question survey was administered to participants to gather information on their backgrounds, technical skills, professional experience, motivations, barriers, program delivery experiences, and outcomes.
- Semi-structured Interviews: Individual interviews were conducted with a subset of participants to gain a deeper understanding of their experiences, motivations, and challenges. A maximum variation sampling strategy was used to ensure diverse perspectives.
The analysis involved both quantitative analysis of survey data and qualitative thematic analysis of interview transcripts, guided by the Knowledge-to-Action (KTA) framework.
Datasets
Not discussed in the paper. The study focuses on evaluating an educational program rather than training or evaluating AI models on specific datasets.
Key Insights and Ideas
- Caregivers are motivated by more than just career advancement: While career shifts were a motivator, participants also expressed a strong desire for personal growth, continuous learning, and combating skill degradation during extended leaves.
- The "identity outside of parenthood" is crucial: For many caregivers, the program offered a way to reconnect with their professional identity and build confidence beyond their caregiving roles, combating feelings of isolation and skill depreciation.
- Accessibility and flexibility are paramount: The C&ML program's success was significantly attributed to its flexible structure, including asynchronous learning, extended TA support outside working hours, and a small cohort size, which directly addressed the unpredictable schedules of caregivers.
- Targeted support can overcome significant barriers: The program's focus on creating a "safe space" for discussing challenges, coupled with instructors and TAs who had lived parenting experience, fostered a supportive environment that mitigated common barriers like imposter syndrome and the rapid pace of AI.
- Confidence is a key outcome: Beyond technical knowledge, the program demonstrably boosted participants' confidence in discussing AI professionally, learning new technical skills, and balancing their roles as caregivers and learners.
Key Results
- High Completion Rate: The program achieved an 86% completion rate among 93 participants.
- Career Transitions: Approximately 20% of participants transitioned into AI-related roles or pursued further education (e.g., graduate studies in AI, new ML-specific roles).
- Skill Application: 27% of participants reported actively applying developed ML skills or discussing ML regularly in their current roles.
- Increased Confidence: Participants showed significant increases in confidence across various AI-related skills, particularly in discussing AI in a professional setting and learning new, unknown concepts.
- Motivations: The most frequently cited skill-specific motivations were understanding ML basics and building comfort discussing/applying ML at work. General motivations included learning a new skill and career advancement.
- Barriers Encountered: Common barriers included the rapid pace of ML, discrimination (sexism), lack of similar role models, and the "motherhood penalty."
- Program Effectiveness: Participants highly valued outside-of-working-hours TA access, small cohort size, and asynchronous course delivery.
Limitations and Future Work
- Self-Selection Bias: Participants voluntarily enrolled, potentially leading to a sample that is more motivated or predisposed to adopting AI concepts than the general caregiver population. Individuals facing significant financial or support challenges might have been unable to participate.
- Researcher Involvement: The evaluation team's involvement in program development could introduce bias. The authors mitigated this by keeping reflexive notes and acknowledging assumptions.
- Program Evolution: The program's content and delivery evolved across cohorts, potentially creating variability in the studied educational experience.
- Future Work: The authors suggest exploring "part two" of the course focusing on practical work experience, internships, or knowledge mobilization. They also emphasize the need for continued efforts in AI education and empowerment for underrepresented groups, highlighting the potential of programs like C&ML to provide equitable access and increase confidence across industries.
Summary updated