Choosing the Right Exam Type for Your Setup
Before creating your first paper in DASES, identify the exam type you are setting up. DASES supports four primary types: Internal Assessment (short tests, 30-50 marks, typically 1-1.5 hours), End-Semester Examination (comprehensive papers, 70-100 marks, 3 hours), Class Test (brief evaluations, 10-20 marks, 30-45 minutes), and Supplementary/Re-Evaluation (processed against the same rubric as the original exam). The exam type selection sets default structural parameters — section requirements, mark ranges, Bloom's distribution targets — that guide the paper setup process. Select the type that most closely matches the exam you are grading.
Configuring the Question Paper: Two Approaches
There are two ways to enter your question paper into DASES. Approach 1 — Upload a PDF: If you have your exam paper as a PDF, upload it in the paper configuration screen. DASES will use OCR to extract the question text, question numbers, mark allocations, and section structure. Review the extracted content for accuracy and make any corrections needed. This approach is fastest when you already have the paper in digital format. Approach 2 — Type directly: Enter each question by hand in the DASES question builder. For each question, specify the marks, select the question type (long answer, short answer, numerical), and note any OR-question pairing. Typing directly gives the most control over the final configuration and is recommended for complex paper structures.
The Critical Step: Writing Effective Model Answers
The quality of your model answers is the most important factor determining the accuracy of AI grading output. A vague or incomplete model answer produces a weak rubric; a detailed, well-structured model answer produces a precise rubric that the AI can apply effectively. For each descriptive question, write the model answer at the level of detail you expect from a top-scoring student. Include every concept, formula, step, or example that should earn marks. Do not summarize — write the full ideal response. For numerical problems, include every step of the working. For essay-type questions, include topic sentences, supporting evidence, and a conclusion. The more complete the model answer, the more accurately DASES can identify which criteria a student's answer meets.
Reviewing and Refining the Auto-Generated Rubric
After you submit each model answer, DASES generates a rubric automatically. The rubric appears as a list of criteria, each with a suggested mark weight. Review this list carefully against your own marking intentions. Verify that the criteria capture the key learning outcomes you intended to test. Check that the mark weights reflect the relative importance of each element. Look for any criteria that seem too granular or too broad and adjust accordingly. Add criteria for elements the AI did not identify — particularly any discipline-specific conventions or course-specific expectations that the AI wouldn't know about. Add "alternative acceptable answers" for criteria where you know students may have been taught different but equally valid approaches.
Scanning Best Practices: Getting Clean Input for the AI
The quality of the scanned answer booklet directly affects AI recognition accuracy. Follow these scanning best practices. Resolution: set your scanner to at least 300 DPI, and 400-600 DPI if students write in pencil (pencil marks require higher contrast). Orientation: ensure pages are scanned in the correct orientation — most scanners auto-detect this, but check the output PDF for any upside-down pages. File size: large PDFs (over 500MB) may cause slow uploads; if your scanner produces large files, reduce DPI slightly or use a PDF compression tool. One file per batch: scan all papers for the same exam into a single PDF if your scanner supports batch scanning — DASES will automatically split the combined PDF into individual student booklets.
The Review Session: A Framework for Efficient Quality Control
The review dashboard is where you add human judgment to the AI's work. Structure your review session systematically. Start with the flagged answers (shown in orange) — these are the cases where the AI had lower confidence, and they deserve your full attention. For each flagged answer, view the student's original handwriting, the AI's transcription, the criterion evaluation, and the proposed score. Decide: is the score appropriate, too high, or too low? Override with a single click if needed. After reviewing all flags, do a spot-check: randomly select 5-10% of the non-flagged answers to confirm the AI is applying the rubric correctly. If you notice a systematic pattern of error in the non-flagged answers, adjust the relevant rubric criterion and re-run the evaluation for that criterion across the batch.
After Publication: Managing Student Queries
After publishing results, expect a small number of student queries — typically 5-15% of the class will have questions about their score. The DASES portal routes student queries to your faculty dashboard, where you see the student's question alongside their answer, the AI evaluation, and the rubric. Most queries can be resolved in under 2 minutes: you either confirm the AI's evaluation with a brief explanation, or override the score if the student's query reveals a genuine evaluation gap. This structured query management is significantly more efficient than ad-hoc WhatsApp or email queries, and it creates an auditable record of every score adjustment decision.