Doctoral research record · Data collection closed
This domain hosted the empirical platform of a doctoral study: a purpose-built Learning Management System in which adult learners studied real courses while the system recorded, with their informed consent, how design decisions about information processes shape motivation and learning success.
The platform has been retired. Until 14.03.2026 this site hosted the data collection and the LMS used in the study. Both have been removed from the server; the collection was completed successfully and the data has been integrated into the dissertation. This page remains as the citable record of the study.
Kind regards,
Lars Arnold Ritter
How can information processes in Learning Management Systems be organised and managed to sustainably enhance motivation and learning success in adult education?
The central question is approached through three subordinate questions:
Working hypothesis: The optimisation of organisational and managerial decisions about information processes inside a learner-centred Learning Management System, when combined with evidence-based pedagogical strategies, significantly increases motivation, engagement and sustained learning outcomes of adult learners compared with conventional configurations.
The study follows a convergent-parallel mixed-methods design: three empirical strands address the same research question through different data types and converge in a formal triangulation step.
Key design figures: 6 months live, 3 empirical strands, 4 matched courses, 18 survey items, 51 active learners, 11 behavioural indicators.
Four real courses ran during the study, matched in pairs on subject, length and structure (eight weeks, twelve lessons in three modules). Two courses had the gamification subsystem switched on (G+: experience points, levels, streaks, achievements), two had it switched off (G−). The pairing isolates the gamification configuration from other course features; the remaining self-selection of learners into courses is addressed statistically (Section 5).
67 adults registered and consented → 51 cleared the activity threshold of at least three completed lessons → 44 completed both survey waves and entered the pre/post analysis pool (26 started in a G+ course, 25 in a G− course).
The empirical work ran on a purpose-built PHP 8.0 application of roughly 7,250 lines of code, developed by the author specifically for this study. Dedicated manager classes (CourseManager, UserManager, QuizManager, GamificationManager, ProgressManager) separate the subsystems along clear domain boundaries, with JSON-based persistence. An off-the-shelf plugin stack was deliberately replaced by this in-house system for four reasons: data sovereignty over learner records, reproducibility of the measurements, freely extensible logging and survey instrumentation, and full GDPR compliance by design.
Six event families were recorded during the study: registrations, lesson completions, quiz attempts (score, duration and per-question correctness), gamification events (every XP gain and achievement with reason and timestamp), streaks, and logins. The survey was implemented as its own lesson type, so responses flowed through the same persistence, pseudonymisation and audit pipeline as the behavioural data. Completing the survey deliberately earned no XP, badges or rank, so that the measurement itself would not distort the motivation it measures.
The study operationalises a three-layer model: design decisions about information processes, the motivational state of the learner, and observable learning outcomes. Each construct receives a primary survey indicator and, where the platform permits, a behavioural counterpart from the log data.
The model additionally specifies a second, direct route from design to outcome: an engagement loop in which feedback events sustain behaviour without passing through reflective self-report. Testing this dual-pathway specification is a core contribution of the dissertation.
| Construct family | Theoretical basis | Items |
|---|---|---|
| Basic psychological needs | Self-Determination Theory (Ryan & Deci; Vansteenkiste et al.) | Autonomy, competence, relatedness, global self-determination |
| Goal orientation, 2×2 | Pintrich | Mastery-approach, mastery-avoidance, performance-approach, performance-avoidance |
| Subjective task value | Expectancy-Value Theory (Wigfield & Eccles) | Intrinsic value, attainment value, utility value, cost (reverse-scored) |
| Interest | DACH interest tradition (Krapp; Renninger & Hidi) | Situational interest (2), individual interest (2) |
| Demographics | Age band, prior digital experience |
„Ich kann meinen Lernweg auf der Plattform selbst gestalten.“ (autonomy)
„Ich möchte den Lerninhalt so gründlich wie möglich verstehen.“ (mastery-approach)
„Das Gelernte ist für meine berufliche oder private Praxis nützlich.“ (utility value)
„Während der Lektionen fesselt mich das jeweilige Thema spontan.“ (situational interest)
| Indicator | Definition | Reads on |
|---|---|---|
max_streak_days | Longest run of consecutive days with at least one interaction | Autonomous engagement |
voluntary_return_days | Distinct days a learner opened a session after first completing a unit | Sustained interest |
fast_completions | Quizzes finished in under two minutes with at least 80 % correct | Perceived competence |
completion_ratio | Lessons completed ÷ lessons total | Completion |
avg_quiz_pct, pct_attempts_above_80 | Mean quiz percentage; share of attempts at or above the 80 % mark | Cognitive outcome |
lesson_abandonment_ratio | Lesson sessions started but not completed ÷ sessions started | Perceived cost |
active_days_span | Span from first to last completion ÷ course duration | Retention |
Survey and behavioural indicators are treated as measurements of the same construct at different timescales: the streak, for example, traces autonomous engagement day by day, while the autonomy item captures its reflective evaluation after the course.
For each construct, the mean indicator value in the gamified pool is tested against the non-gamified pool with Welch's t-test, which tolerates unequal variances and group sizes:
t = ( x̄G+ − x̄G− ) ⁄ √ s1²/n1 + s2²/n2
Effect sizes are reported as Cohen's d with bootstrap 95 % confidence intervals, interpreted against the conventions of the gamification literature (|d| ≥ 0.20 small, ≥ 0.50 moderate, ≥ 0.80 large). Because the construct tests are related, multiple comparisons are controlled with the Benjamini-Hochberg false-discovery-rate procedure:
reject all H(1) … H(k), where k = max { i : p(i) ≤ (i/m)·q }, q = 0.05 ordered p-values p(1) ≤ … ≤ p(m) across the m construct-level tests
For each of the sixteen Likert items, the per-learner difference (post minus pre) is tested against zero with a paired-samples t-test and standardised as Cohen's dz:
dz = Δ̄ ⁄ sΔ with Δ = xpost − xpre
A difference-in-differences contrast then asks whether gamification moderates the change: mean change in G+ against mean change in G−, again with Welch's t-test and Cohen's d.
The model's central prediction, that design reaches outcomes through motivational states, is tested with the regression-based mediation procedure standard in educational research (Hayes):
M = i1 + a·X + e1 Y = i2 + c′·X + b·M + e2 X: design configuration · M: motivational mediator · Y: outcome · indirect effect a·b, direct effect c′, total effect c = c′ + a·b
Indirect effects are estimated by bootstrap resampling with 5,000 iterations, separately for each motivational mediator. To check stability against learners self-selecting into courses, propensity-score weighting on the demographic indicators and prior digital experience is applied.
Internal consistency of the survey composites is assessed with Cronbach's alpha (source scales report .70 to .85):
α = k/(k−1) · ( 1 − Σsi² ⁄ sX² )
Finally, each construct is triangulated across the three strands: convergence when literature, survey and behavioural record support the same claim, partial convergence at two of three, divergence otherwise. The triangulation pattern across the construct space is the principal evidentiary product of the study.
The empirical work ran under a research-ethics regime combining the GDPR, the institutional research-conduct expectations of ULSIT and a published technical-privacy framework. Five procedures carried it in practice:
Withheld until the defence. The findings are deliberately not published on this page. The dissertation is under evaluation and the results will be presented at the public defence in Sofia.
At headline level only: the behavioural record responded more strongly to the design configuration than the self-reports did, and the effect of gamification on completion appears to operate primarily through a direct engagement loop rather than through reflectively reported motivational states. Full tables, effect sizes and the complete analytic pipeline are documented in the dissertation.
The manuscript, the survey instrument (German operational wording), the codebook for the behavioural and survey variables, and aggregate analytic outputs can be made available for inspection on request, given a legitimate academic interest. Pseudonymised raw data is shared strictly within the consent terms agreed with the participants.
Requests: via LinkedIn, Lars Ritter.