What does generative AI do to adult learners’ wellbeing? A Ghana study found useful support alongside strain and mistrust
Among 212 adult learners, overall digital-wellbeing ratings sat above the scale midpoint, yet minorities reported disrupted sleep, discomfort and less exercise. Six selected participants described both relief and frustration; the one-setting, self-reported study cannot show that AI caused either outcome.
The short version
- What happened
- The cross-sectional survey included 212 adult learners; 191, or 90.1%, reported current GenAI use. Ninety-nine per cent had smartphone access and 80.2% most often accessed GenAI through a phone.
- Who it affects
- Adult learners balancing study, work and family may value quick explanations and drafting help while also encountering misinformation, frustration and added checking work.
- What remains unproven
- The survey was cross-sectional and self-reported, so it cannot establish that GenAI caused any wellbeing or quality-of-life outcome.
Keep in mind
The survey was cross-sectional and self-reported, so it cannot establish that GenAI caused any wellbeing or quality-of-life outcome.
The 212 respondents came from one Ghanaian higher-education setting and are not a representative sample of all adult learners in Ghana or elsewhere.
The six qualitative participants were purposively selected from low scorers who volunteered contact details; their accounts cannot estimate prevalence.
Several quality-of-life items had only 111–135 substantive responses because non-substantive options were excluded from scoring.
Social desirability, the group setting and researcher interpretation may have influenced the qualitative material.
Source analysis · 1 linked source. Primary analysed source date: 6 November 2026.
Read the full report · 9 min The evidence, methods, implications and limitations
The answer: learners described a mixed experience, not a simple wellbeing gain or loss
A peer-reviewed study at one Ghanaian higher-education setting found that adult learners generally rated their digital wellbeing while using generative AI above the midpoint of a seven-point scale. They associated the tools with problem-solving, productivity, interest in new topics and clearer thinking. At the same time, the survey recorded a smaller group reporting sleep disruption, changes in eating, physical discomfort or reduced exercise. Six selected participants then described how assistance with studying, work and personal questions could reduce pressure, while wrong or unhelpful answers produced frustration, mistrust and concern about overdependence.
The evidence does not support a claim that GenAI improved or harmed wellbeing overall. The survey measured perceptions once, did not compare users with a randomly assigned non-user group and did not track sleep, health, grades or work performance objectively. The qualitative phase deliberately invited volunteers with lower survey scores, so their stories explore difficult experiences rather than estimate how common those experiences are. The useful conclusion is narrower: adult learners can experience efficiency and emotional support at the same time as strain, and education policy needs to make room for both rather than treating use or non-use as the outcome.[1]
Who took part, how they used GenAI and how the study was designed
The quantitative phase recruited 212 adult learners from a Ghanaian university context between August and October 2025. The paper defines adult learners broadly as people aged 18 or older enrolled in distance or higher education, a group often balancing study with employment and family responsibilities. Of the 212 respondents, 191 reported current use of generative-AI tools and 21 did not. Smartphone access was nearly universal at 99%, 71% had a laptop and 80.2% said a smartphone was their main access device. Education and research were the most common stated purpose, reported by 76.4% of respondents.
Usage intensity varied. About 31.6% reported daily use, 27.4% weekly use and 41% rare use; slightly more than half spent less than an hour a day with the tools. Almost half reported occasional internet instability affecting access and 27.8% reported power outages. The questionnaire used established digital-wellbeing and quality-of-life measures, adapted and pre-tested with another adult-learner cohort. Negative physical items were reverse-coded so higher scores consistently represented better wellbeing. Reliability was assessed, and the quality-of-life scale had an internal-consistency value of 0.829. These are perception measures, not clinical diagnoses or behavioural telemetry.[1]
The distinction that matters
How to read the two parts of the study
The survey describes patterns across 212 learners; the six-person conversation explains selected experiences. The two parts answer different questions.
Survey
212 adult learners
Self-reported averages and proportions
Current users
191 respondents
90.1% of the survey sample
Story phase
6 selected volunteers
Detailed low-score experiences
Not established
GenAI use
Caused wellbeing gains or harms
What the numbers say—and why averages conceal important minorities
The overall digital-wellbeing mean was 4.58 on the seven-point scale, with a standard deviation of 0.76. The mental subscale averaged 4.09, the emotional subscale 4.51 and the reverse-coded physical subscale 5.14. Stronger individual ratings included problem-solving, interest in new things, productivity and decisiveness. The quality-of-life composite averaged 2.06 on a three-point substantive scale, suggesting modest perceived gains in control, activity and time efficiency among those providing scorable answers. The paper itself cautions that several quality-of-life items had only 111 to 135 substantive responses because “does not apply” and “do not know” were excluded.
The averages should not erase the tails of the distribution. Seventeen per cent reported disturbed sleep, 16.5% physical discomfort, 16.5% less exercise and 10.9% changed eating habits associated with their use. Those figures are not medical incidence rates, because respondents attributed effects themselves and no baseline or clinical assessment was available. Age-group comparisons were not statistically significant across mental, emotional, physical or overall wellbeing and quality of life. That non-significance also should not be turned into proof that age never matters: a sample of 212 at one setting may lack power for smaller differences and may not represent older learners elsewhere.[1]
What six selected stories add to the survey—and what they cannot represent
After the survey, researchers purposively invited six participants who had lower digital-wellbeing or quality-of-life scores and had volunteered contact details. Three men and three women aged 20 to 40 joined a 45-to-60-minute group conversation in November 2025. The team recorded audio and video, retained Ghanaian-language expressions, shared reconstructed stories back with participants and used an African oral traditional storytelling framework alongside inductive thematic analysis. This approach gives cultural and emotional context to a numerical scale and makes the participants collaborators in interpreting their accounts.
The accounts described quick explanations, vocabulary support, drafting help and reduced anxiety when time was short. They also described disappointment when summaries, calculations or image tasks failed, anxiety about inaccurate information and worries that children or students might substitute generated answers for reading and judgment. Participants reported coping by lowering the tool’s authority, checking outputs, returning to peers or lecturers, taking breaks and treating frustration as a prompt to learn. Because the six were selected for low scores and spoke in a shared setting, their experiences should not be presented as typical of all Ghanaian learners. Social desirability and the researchers’ presence may also have shaped what was said.[1]
What institutions can test without turning personal coping into a substitute for support
For universities and distance-learning providers, the study points to a practical question: are learners being given a tool, or a supported way to use it? Clear rules about acceptable assistance, privacy, checking sources and escalating errors can reduce ambiguity. Courses can show students how to compare generated claims with primary material and how to keep human discussion in the learning process. Mobile-first design matters in a setting where smartphones dominate access, while unreliable connectivity and power make cloud-only support uneven. Counselling, academic advising and disability services should remain reachable without requiring a learner to disclose sensitive information to a chatbot.
Those proposals still need evaluation. Institutions should measure whether guidance changes verification behaviour, workload, sleep, stress, assessment quality or course completion, rather than assuming a short workshop solves technostress. Staff should distinguish a learner choosing an efficient drafting aid from a learner depending on a system because other support is unavailable. Personal strategies such as breaks or peer discussion can help, but they should not transfer responsibility for platform errors, confusing assessment rules or weak student services onto individuals. The study offers hypotheses and lived context; it does not test an institutional intervention.[1]
What would change the assessment
Stronger evidence would follow adult learners over time, record when and why they use particular systems and compare changes in wellbeing with matched non-use periods or carefully designed interventions. Objective or validated outcomes could include sleep measures, study time, assessment results, retention, stress screening and help-seeking, with safeguards for sensitive data. A larger qualitative sample drawn across several institutions, regions, ages, languages and levels of connectivity would test whether the reported tensions recur or reflect the particular group and setting. Analyses should also separate educational use from work, health and family questions because the risks differ.
For now, the defensible reading is that these learners did not report one uniform effect. Most survey respondents described broadly positive digital wellbeing, while meaningful minorities reported physical strain and the six selected collaborators described both support and frustration. The paper broadens the evidence beyond wealthy English-speaking campuses and foregrounds adult learners juggling multiple responsibilities. It does not prove that GenAI causes better quality of life, anxiety, sleep loss or dependency. The assessment would change if longitudinal, multisite evidence showed consistent effects and if tested support measures improved outcomes without sacrificing learning, privacy or human connection.[1]
Keep following the evidence
Join The Impact Brief founding list for selected AI reports, their effect on people and the evidence limits. Read a sample first.
Continue the story
Read next
Continue exploring higher education: compare the finding and its evidence limits.
What do student prompts reveal about agency? The task phase mattered more than the deadline
A University of Sydney preprint classified 962 prompts from 105 first-year students. Most showed some student contribution, but explicit constraints were rare and self-reported journals cannot establish learning or the true prevalence of delegation.
Read the next reportExplore the full source and evidence record
What this means for people
- Adult learners balancing study, work and family may value quick explanations and drafting help while also encountering misinformation, frustration and added checking work.
- Universities can make human academic and wellbeing support visible, design for mobile and intermittent connectivity, and teach verification without assuming that tool availability equals benefit.
- Learners should not read the study as a medical assessment of their own sleep or mental health; its outcomes are self-reported perceptions from one educational setting.
Global context
Adult and distance learners in Ghana often reach digital services through smartphones and may face intermittent power or connectivity. That makes this study a useful counterweight to evidence from continuously connected residential campuses. Its results still require replication across African regions and in other institutional, linguistic and economic settings before they can support broad claims.
What the evidence does not yet show
- The survey was cross-sectional and self-reported, so it cannot establish that GenAI caused any wellbeing or quality-of-life outcome.
- The 212 respondents came from one Ghanaian higher-education setting and are not a representative sample of all adult learners in Ghana or elsewhere.
- The six qualitative participants were purposively selected from low scorers who volunteered contact details; their accounts cannot estimate prevalence.
- Several quality-of-life items had only 111–135 substantive responses because non-substantive options were excluded from scoring.
- Social desirability, the group setting and researcher interpretation may have influenced the qualitative material.
What to watch next
- Longitudinal studies separating educational, work, health and personal GenAI use.
- Multisite African evidence with larger and more diverse adult-learner samples.
- Trials of verification teaching, clear assessment rules and accessible human support.
- Objective learning, sleep, workload and retention outcomes alongside self-report.
- Whether mobile-first and low-connectivity design reduces inequity without increasing privacy risk.
Living evidence record
Impact record IAI-082HHQ9
Evidence stage
Studied
Confidence
Supported
Reporting basis
Source analysis
Independent or research support
Present
Record status
Monitoring
Last checked
6 November 2026
Source trail
1 direct source across 1 source type.
People impact
Documented in this record.
Uncertainty
Limits and next checks are explicit.
Stages describe the evidence available—not whether a technology is good or bad. See the public method.
Single-source reporting disclosure
This record analyses one direct source. It can establish what Scientific Reports published or reported, but it is not independent corroboration of every performance claim or predicted outcome. The confidence label will change only when broader evidence is added.
Evidence trail
Sources used for this report
Links checked 6 November 2026
Research question, topics and regions
Digital wellbeing, quality of life and coping among adult learners using generative AI in a Ghanaian higher-education setting
Topics: Adult learners · Digital wellbeing · Higher education · Generative AI · Technostress · Mixed methods
Regions: Ghana · Africa · Global
This report is labelled source analysis. We summarise and analyse source material in our own words; company statements remain attributed claims until independently supported. Translated summaries preserve the meaning of the original source and link back to it. Read our editorial standards.
Reader commentary
What does this mean where you are?
Tell us what you have seen at work, in your community or in everyday life. A question is welcome too. No account is needed; we review notes before publishing them.
Explore commentary across the portal →Published reader notes
0No published reader notes yet. You can start the evidence-led discussion above.
Prefer a private correction or response? Contact the newsroom.