aiboxlab/edu-interactions

🎓 IV WAPLA: EDU AI Assistant - Digital Learning Insights

The IV Workshop on Practical Applications of Learning Analytics and Artificial Intelligence in Brazil (Workshop de Aplicações Práticas de Learning Analytics em Instituições de Ensino no Brasil, WAPLA 2026) is a satellite event of the XV Brazilian Congress on Informatics in Education (Congresso Brasileiro de Informática na Educação, CBIE 2026).

In 2026 the 4th Edition of WAPLA, in partnership with the Cogna Learning Engineering Core (Núcleo de Engenharia do Aprendizado Cogna (NEAC)), promotes two educational data analysis competitions. The first is the Interações do EDU (EDU Interactions) competition.

For more information, refer to the official WAPLA website: WAPLA 2026

📚 Overview and Context

This dataset contains academic records and interaction logs from students enrolled in undergraduate distance learning courses who had access to EDU, an AI-based generative learning assistant integrated into the institution's virtual learning environment. EDU is used during various learning activities — such as reading instructional materials, viewing educational videos, and completing formative exercises — to answer questions, provide explanations, suggest prompts, summarize content, and support students throughout their learning process.

---

🗂️ Dataset Structure

The dataset is composed of two relational tables that can be joined by id_aluno and disciplina:

Table	Description
GRADE	Academic results at the student-discipline level
INTERACTIONS	EDU usage records in learning activities

Table: GRADE

Variable	Type	Description
id_aluno	Integer	Anonymous student identifier
disciplina	Text	Discipline/subject name
inicio_disciplina	Date	Discipline/subject start date
termino_disciplina	Date	Discipline/subject end date
status_disciplina	Categorical	Final enrollment status (e.g., PASSED, FAILED, CANCELED)
nota_final	Numeric	Final grade obtained in the subject (scale 0–100)
is_eletiva	Boolean	Elective discipline (TRUE) or required (FALSE)

Observations:

• The same student may appear in multiple records due to enrollment in different disciplines;
• Cancelled disciplines may also be present;
• Discipline durations vary between offerings;

Table: INTERACTIONS

Variable	Type	Description
id_aluno	Integer	Anonymous student identifier
momento	Categorical	Learning context (LEARNING_VIDEO, LEARNING_TEXT, LEARNING_QUESTION)
disciplina	Text	Discipline where the interaction occurred
qtd_mensagenstotal	Integer	Total number of messages exchanged
qtd_mensagenserro	Integer	Messages with system errors or failed responses
qtd_mensagensedu	Integer	Messages generated by EDU
qtd_mensagensaluno	Integer	Messages sent by the student
qtd_csat_respondido	Integer	CSAT (Customer Satisfaction) evaluations responded to
csat_medio_respostas	Numeric	Average value of CSAT evaluations (missing values possible)
timestamp_inicio_mensagens	Date	Start date of the interaction record

Learning Context (momento):

• LEARNING_VIDEO are interactions during educational video activities;
• LEARNING_TEXT are interactions during textual content reading;
• LEARNING_QUESTION are interactions during exercises, quizzes, or question-based activities;

Satisfaction Variables: Not all interaction records contain satisfaction evaluations. The variables qtd_csat_respondido and csat_medio_respostas register this information when available.

---

Dataset provided for educational and research purposes. The competition is hosted on Kaggle: Interações do EDU