Judging

Every project is scored out of 100 points on seven criteria. The same rubric is used in Round 1 and at the finale.

How a score is worked out

For each criterion, the judge chooses a level from 1 to 4. Half levels (1.5, 2.5, 3.5) are allowed.

Level 1 earns 25% of the criterion's points, level 2 earns 50%, level 3 earns 75% and level 4 earns 100%.
For example, level 3 on criterion 1 (20 points) earns 15 points.

Problem and relevance. Is the problem real, important to the community and clearly explained, with evidence from the people affected?
Use and understanding of AI. Does the team use AI where it genuinely helps, and can they explain how it works and where it falls short?
Creativity and originality. Is the idea fresh, or a new take on an existing solution that suits Zanzibar?
Technical execution. Does the prototype work? Was it tested, and improved after testing?
Responsible AI and ethics. Has the team thought about privacy, fairness, safety and mistakes, and been honest in the AI Use Log?
Impact and feasibility. Could it make a real difference, and could it be used with the resources, devices and connectivity people actually have?
Communication and teamwork. Are the brief and video clear, and did every team member contribute?

Judges see detailed descriptions of each level when they score.

Judges score content, not polish or language. Kiswahili and low-resource formats, such as a hand-drawn design or a phone video, are equally valid.
Judges do not see the school or island of the projects they score.
Judges declare any conflict of interest, for example a family member or their own school, and never score those projects.
Projects near the cut-off get a second judge. If two scores differ by more than 15 points, a third judge scores the project.
The judging lead reviews scores across judges so no project is disadvantaged by a strict or generous judge.