Updated: 01 Apr 2026, 23:25+02:00

Assessment Logic and Generative AI Use Policy

Table of contents

Assessment Logic
Generative AI Use Policy

Assessment Logic

Assessment consists of four elements:

Assessment type	Examination	Team Points	Individual Points	Totals
Gate to examination	Assignment	n/a	n/a	n/a
Examination	First Submission	30	20	50
Examination	Oral Examination	5	30	35
Examination	Final Submission	15	–	15
	Totals	50	50	100

Rubric: Assignment (ungraded gate to examination)

Category	Criteria
Team Composition	Team name and list of contributors Team repository that examiner can “git pull” (cannot be changed later) Meta-goals per contributor: target grade and personal goals
Value Proposition	Target user with a plausible problem and team’s (app-based) solution proposal App must implement a two-sided platform
Target Scope	Visual scoping of Web App, primarily via scribbles of UI screens

Notes

Achievement of “Value Proposition” and “Target Scope” criteria are assessed via an in-class peer review format. Informed by these results, the Lecturer decides “pass / rework / fail”. Rework due date is announced by Lecturer.

Students must pass the mandatory Assignment (“Studienleistung”) to be eligible for subsequent examination.

Rubric: First Submission

Category	Team	Individual	Points + Criteria
Product Discovery	10	–	- Evidence folder with “raw material” aimed to (1) define design challenge, (2) understand target user with their core problem, (3) propose solution elements, and (4) test ideas - Value Proposition: what is the problem? Who has it? How to solve it with an App?
Product Delivery	20	–	- Happy Path derived from Value Proposition (clear priorities) - After “git clone”: within 10 min, examiner can reproduce App Happy Path locally by following README.md instructions (max. 6: screen recording of Happy Path as fallback) - App fulfils all Mandatory Requirements - Data Model visualized and described, must match implementation
Individual Contribution	–	20	- Individual contributions listed and backed up by proof (primarily through exemplary git commit traces) - Top 3 contributions explained: my contribution, why I am proud of it, which challenge I overcame - At least 2 relevant Design Decisions in the mandatory format, with proof of regarded options (tangible self-created artifacts like repo branches, not list of links or similar) - Contributed to app source code
Totals	30	20	50

Mandatory Requirements and Forbidden Technology

Mandatory Requirements are checked with First Submission, see rubric above.

Forbidden Technology is checked with First Submission, and again with Final Submission.

Accidental inclusion of Forbidden Technology may be remedied after First Submission, within 2 weeks of Lecturer notice.
If any Forbidden Technology is detected with Final Submission, it is automatically awarded 0 Points (out of 15).
The Team is fully accountable for not including any Forbidden Technology, also if Lecturer doesn’t catch it with First Submission, but notices it in Final Submission.

Mandatory Requirement	Forbidden Technology
Written in Python	Any (even minor) use of JavaScript / TypeScript; exception: JS functionality bundled with Bootstrap
Use of Flask	Replacing Flask with another web app framework, e.g., FastAPI
Use of Jinja2	Replacing Jinja2 with another templating engine, e.g., Mako
App handles multiple HTTP requests with varied business logic (no content website)	–
SQLite, with exactly one database file checked into repository	Any other database technology (including NoSQL databases), or remote hosting of database
User roles, including authorization flow(s)	–
At least 1 “headless” API that delivers a JSON file	–
Must be executable natively on a current Windows or MacOS system	Packaging any part of your App in a Docker container, a Virtual Machine, or similar

Rubric: Oral Examination

Category	Team	Individual	Points + Criteria
Goal Setting	5	–	- Path from First to Final Submission, with reasoning how it is aligned with (1) Value Proposition and (2) current state of App
Design Insight	–	10	- When asked, able to logically / reasonably explain particular decision(s) in own area of responsibility - Shows overall intimacy with problem space, target user, proposed solution
Code Insight	–	15	- Demonstrates technical mastery; signals: (1) easily navigates in codebase; (2) answers are convincing (not wrong/ reasonable) and to-the-point; (3) can assess consequences of proposed changes
Communication	–	5	- Stays within time budget; shows conversational ability under pressure (remains calm/ professional in examination setting)
Totals	5	30	35

Rubric: Final Submission

Category	Team	Individual	Points + Criteria
Consistency	10	–	- Goal achievement since Oral Examination assessed - It is easy for examiner to see how App is faithful to Value Proposition - No part of the Final Submission contains inconsistencies
Submission Usability	5	–	- Git repo without clutter (e.g., no venv/ folder) - Documentation on publicly accessible website and automatically built from .md files (e.g., via GitHub Pages) - After “git pull”: examiner can easily run functional App by following README.md instructions
Totals	15	–

Generative AI Use Policy

Principle

The goal of this Module is to assess your understanding and ability to create, not the output quality of a tool. You must maintain human agency over every line of code and every sentence that you submit.

Permitted Tools

You are allowed to use Generative AI for assistance. You are fully responsible for any result that you submit, you must understand and be able to explain it. Permitted tools:

Chat-based interfaces (e.g., ChatGPT, Perplexity), e.g., for brainstorming, to learn about concepts and technologies, to explain encountered errors.
IDE code completion (e.g., Copilot within Visual Studio Code) for small snippets or boilerplate code.

Prohibited Tools

The use of Agentic AI is FORBIDDEN. This includes any AI-based tool that plans and executes several steps autonomously, modifies files, manages repositories, generates code structure, etc. without step-by-step intervention by you. This includes the creation of git commits (and commit messages) on your behalf.

Exemplary tools that are forbidden include, but are not limited to: Aider, Antigravity, Claude Code, Codex, Copilot Chat Agents, Cursor, OpenClaw, OpenCode, Replit.

Disclosure

You must maintain a comprehensive AI Directory, as per FB1 Regulations on Generative AI Use. “Catch-all” disclosure (like “AI Tool used for bugfixing”) is generally not sufficient.

Again, any use of Agentic AI is forbidden.

Verification and Sanctions

Your ability explain the artifacts you created (code and documentation) is the primary verification mechanism.

Oral Examination: Your individual grade depends on your ability to explain your own results.
Undisclosed Tool Use: If detected at any point, (e.g., Oral Examination), your Individual Points are capped at 10 (out of 50).
Forbidden Tool Use: If detected at any point, you will fail the Module for Misconduct (“Täuschung”).