Evidence, testing and evaluation - Computer Science IB Study Notes

Evidence in Computer Science refers to documented proof that a solution meets its specified requirements and functions correctly. This includes test plans, test data, screenshots, user feedback, and evaluation reports.

Testing is the systematic process of executing a program with the intent to find errors and verify functionality. Key testing types include:

• Alpha testing: Internal testing by developers before release • Beta testing: External testing by selected users in real-world conditions • Acceptance testing: Final validation that software meets user requirements • White-box testing: Testing internal structures/workings (code-based) • Black-box testing: Testing functionality without knowing internal code

Test data categories are crucial: • Normal data: Valid inputs within expected range (e.g., age = 25) • Boundary data: Values at the limits of acceptable ranges (e.g., age = 0, 120) • Erroneous data: Invalid inputs that should be rejected (e.g., age = -5, "abc")

Evaluation is the critical analysis of how well a solution meets its objectives, considering efficiency, usability, maintainability, and reliability. The evaluation compares actual outcomes against success criteria defined in the requirements specification.

Cambridge Key Term: A test plan is a structured document specifying what will be tested, what test data will be used, expected outcomes, and actual outcomes.

Iterative refinement means using test results and evaluation feedback to improve the solution through multiple development cycles—a fundamental principle in agile methodologies.

Think of testing like a quality control process in manufacturing. Just as a car manufacturer tests vehicles under various conditions (normal driving, extreme weather, crash scenarios), software developers must test programs with different data types.

Real-world application: When Instagram develops a new feature, they follow rigorous testing protocols:

1. Alpha testing: Engineers test the photo filter internally, checking it works with various image formats (JPG, PNG, HEIC)
2. Beta testing: Selected users worldwide test the feature, revealing issues like slow loading on older phones or cultural appropriateness concerns
3. Boundary testing: Images at 0KB (empty file) and maximum file size limits are tested
4. Erroneous data: Non-image files (PDFs, videos) are submitted to ensure proper error handling

Evaluation in context: After launch, Instagram evaluates success using metrics: • Efficiency: Does the filter apply in under 2 seconds? • Usability: Can 90% of users apply it without help? • Reliability: Does it crash less than 0.01% of the time?

Medical device analogy: A blood glucose monitor undergoes extensive testing—normal readings (4-7 mmol/L), boundaries (3.0, 20.0), and erroneous inputs (negative values). The consequences of poor testing here could be life-threatening, highlighting why systematic testing isn't optional.

Mnemonic for test data: N.B.E. = Normal, Boundary, Erroneous—test all three categories every time!