THAProf. Conte's Course AssistantProfessor-controlled student communication

Prof. Dr. Conte's trusted course knowledge

Your question deserves a verified answer.

Write to Prof. Dr. Conte directly inside this app. Your question enters his protected inbox, where he can review it and publish an approved answer.

✓ Stay inside the app✓ Approved by Prof. Dr. Conte✓ Source-grounded
01

Write to Prof. Dr. Conte

Fields marked * are required

Verified course knowledge

Prof. Dr. Conte's answers

Search answers personally reviewed and published by Prof. Dr. Conte.

⌕
Loading published answers…

How Prof. Dr. Conte's assistant works

One inbox. Three channels. Prof. Dr. Conte stays in control.

His students ask on the web or through an approved WhatsApp number. The university-hosted system retrieves only from his trusted course notes, prepares a draft, and waits for his review.

W

Student app

Students write structured questions and remain inside the app.

Working now
IN

Professor inbox

Every submitted question appears immediately in Prof. Conte's protected dashboard.

KB

Course knowledge

Markdown and Obsidian notes stay on the university-controlled server.

Working now
  1. 01

    Student asks

    The message enters the same professor inbox with its original channel.

  2. 02

    Trusted material is retrieved

    The system cites the course files it used and abstains when evidence is insufficient.

  3. 03

    A draft is prepared

    The professor sees the question, retrieved sources, and an editable draft.

  4. 04

    Professor reviews

    The professor edits, rejects, archives, or explicitly approves the answer.

  5. 05

    Answer returns

    Approved answers can enter the FAQ and, for WhatsApp questions, return to the same chat.

Prof. Dr. Fiorentino Valerio Conte

Built for Prof. Dr. Conte

His course communication, in one place.

His dashboard shows evidence, channel, delivery state, WhatsApp connection, and audit history. Profile information and portrait come from the official THA profile.

August 2026 research sprint

Evidence before automation.

The THA Q&A Hub is evaluated as a privacy-aware, human-in-the-loop retrieval assistant—not an autonomous professor. Every draft remains reviewable and unpublished until a professor approves it.

Current conclusionPromising prototype

Technical feasibility demonstrated. Human effectiveness, legal readiness, and production WhatsApp reliability still require institutional evaluation.

Research question

Can verified retrieval reduce repetitive course communication without reducing professor control?

The study separates retrieval quality, unsupported-question abstention, answer grounding, human editing effort, usability, privacy, and operational risk.

Source Hit@388.9%Original lexical baseline
Safe abstention16.7%Original baseline exposed a major risk
Guarded Hit@3100%Small development benchmark
Guarded abstention100%Needs independent validation
01

Key finding

Retrieval presence is not evidence sufficiency.

The original keyword retriever found superficially related passages for five of six unsupported questions. A documented evidence threshold and transparent synonym handling eliminated those failures on the 15-case development benchmark. Because the intervention followed baseline inspection, the perfect guarded result is not proof of generalisation.

What is proven

  • Local course-document retrieval is reproducible.
  • Sources remain visible to the professor.
  • Unsupported questions trigger explicit abstention.
  • Publishing and WhatsApp delivery require professor action.
  • All current application safety tests pass.

What is not proven

  • No real professor workload study has occurred.
  • No student usability study has occurred.
  • No claim of DSGVO compliance is made.
  • No production OpenWA bridge has been fault-tested.
  • The synthetic benchmark is not independent evidence.

Research framework

1 Govern
scope, roles, approval→2 Map
data and failure paths→3 Measure
retrieval, safety, utility→4 Manage
controls and incidents

Grounded in primary RAG research and authoritative GDPR, EU AI Act, EDPB, NIST AI RMF, and OWASP guidance. Legal classification remains the university’s responsibility.