AIMech AI
Home / Curated / Preprocessing Methods / MOOCdb Four-Mode Learner Interaction Schema and Multi-Layer Privacy Partitioning

MOOCdb Four-Mode Learner Interaction Schema and Multi-Layer Privacy Partitioning

Preprocessing Methods AIMech:MOOCdbFourModeLearnerInteractionSchemaAndMultiLayerPrivacyPartitioning DRAFT
Aliases: MOOCdb Standardized Relational Learner Telemetry Schema, MOOCdb Four-Mode Behavioral Interaction and Three-Layer User Anonymization Framework
A cross-platform data standardization, feature-engineering, and privacy-partitioning preprocessing method for Massive Open Online Course (MOOC) behavioral telemetry that transforms heterogeneous raw logs into four relational interaction mode sub-schemas — observing (many-to-many URL-to-URI resource browsing events), submitting (tree-structured problem hierarchies, multi-attempt submissions, and decoupled multi-grader [0,1]-normalized assessments), collaborating (self-referential forum/wiki interaction trees), and feedback (resource-linked survey responses) — while decomposing learner identifiers across a three-layer hierarchy (global_user_id -> course_user_id -> four mode-specific user IDs) that defines six access-control data partitions.
Identity Notes & Concept Scope No AIO class represents MOOCdb's four-mode behavioral telemetry relational schema or three-layer privacy partitioning; DataPreparation (AIO:DataPreparation) is the closest broad match covering structured preparation, normalization, and anonymization of raw interaction logs for machine learning.

Claims & Grounded Evidence

4 Claims
Mechanism #four-mode-relational-telemetry-decomposition-and-hierarchical-tree-encoding
MOOCdb normalizes learner interactions across Coursera and edX into four lossless transactional sub-schemas (Section 4.1-4.4, Figures 3-8, 17): (1) observing mode, which decouples webpage container URLs (`urls`) from embedded resource URIs (`resources`) via a many-to-many `resources_urls` join table; (2) submitting mode, which encodes arbitrary module-problem-subproblem hierarchies as trees via `(problem_id, problem_parent_id, order_id)` (Figure 4) and decouples `submissions` from multi-grader [0, 1]-normalized `assessments` (Figure 5); (3) collaborating mode, which stores forum threads, votes, and wiki edits as per-student transactions with self-referential `collaboration_parent_id` links (Figures 6-7); and (4) feedback mode (`surveys`, `questions`, `answers`, `feedbacks`, Figure 8).
Mechanism #three-layer-localized-user-id-anonymization-and-six-partition-access-control
To control cross-course and intra-course linkability without mutating interaction tables, MOOCdb separates `user_pii` (`global_user_id`) from `global_user_table` (`global_user_id` -> per-course `course_user_id`) and `course_user_table` (`course_user_id` -> four distinct mode identifiers `observing_user_id`, `submitting_user_id`, `collaborating_user_id`, `feedback_user_id`, Figures 9-10), inducing six access-control partitions (Multi-course, Single-course, and Table/Mode-level, each with or without the `collaborations` table, Figure 11).
Application #mitx-6002x-tenfold-compression-mooc-en-images-cube-and-bkt-irt-feature-foundry-application
Applied to MITx 6.002x (154,763 registrants, 132,286,335 observed events in 6,120 MB and 6,313,050 submissions in 341 MB), MOOCdb reduces storage by over 10x relative to a 70 GB name-value transcription and accelerates behavioral queries by ~10x, powering the MOOC En Images 3-axis analytics framework (Time x Student Cohorts x Space, Figures 13-14) and standardized exports for Bayesian Knowledge Tracing (BKT) and Item Response Theory (IRT).
Limitation #public-forum-timestamp-correlation-and-cross-course-profile-reidentification-limitation
Even when `user_pii` and `course_user_table` are withheld (Partition 4, Mode Table access), releasing the `collaborations` table alongside `observed_events` permits re-identification of forum-posting learners via timestamp correlation between forum post timestamps and forum-visit events in `observed_events`; similarly, multi-course access without `collaborations` (Partition 3) remains vulnerable to profile-intersection de-identification when learners publicly disclose course combinations and grades (Example 5.1, pp. 14-17).

⚠️ Curated Limitations & Negative Results

Artificial Intelligence Ontology (AIO) Grounding

Mapped AIO Classes
AIO Target Class Mapping Relation Mapping Rationale
AIO:DataPreparation
Data Preparation
 broad_match MOOCdb prepares and transforms raw JSON/XML/SQL interaction streams into normalized relational tables and privacy-controlled data partitions for feature extraction and statistical modeling.

Evidence Sources & Primary Literature

3 Sources
Source ID Publication / Document Type / Access Reference & Checksum
r840-primary-pdf
Kalyan Veeramachaneni, Sherif Halawa, Franck Dernoncourt, Una-May O'Reilly, Colin Taylor, Chuong B. Do
primary paper
full_text
ID: https://arxiv.org/abs/1406.2015
sha256:453e8fca5ac0…
r840-arxiv-abs
Kalyan Veeramachaneni, Sherif Halawa, Franck Dernoncourt, Una-May O'Reilly, Colin Taylor, Chuong B. Do
primary paper
abstract_only
ID: https://arxiv.org/abs/1406.2015
sha256:4ce0e034311f…
r840-metadata-cr
Kalyan Veeramachaneni, Sherif Halawa, Franck Dernoncourt, Una-May O'Reilly, Colin Taylor, Chuong B. Do
primary paper
abstract_only
ID: https://arxiv.org/abs/1406.2015
sha256:b32fc1a1d422…

Ontology & Grounding Topology

Open in interactive network graph →
CONCEPT KIND Preprocessing Methods AIMech:MOOCdbFourModeLearnerInteractionSchemaAndMultiLayerPrivacyPartitioning MOOCdb Four-Mode... 4 claims · 3 sources AIO GROUNDING AIO:DataPreparation (1)

Curation Provenance

Curator: Antigravity
Method: ai_assisted
Model: Gemini
Created: 2026-10-19T15:23:00Z
Updated: 2026-10-19T15:23:00Z
Source File: data/curated/MOOCdbFourModeLearnerInteractionSchemaAndMultiLayerPrivacyPartitioning.yaml · Schema: curated_concept.yaml

Explore across X-Mech

Related computational & biological Mechs: Embedding Map Network Graph ProteinTraitsMech (ESM-2 pLM) DUFMech (Domain pLM) TraitMech CultureMech All 14 X-Mech Repositories