Four Speakers, One Microphone
You are rating a four-candidate group discussion: eight minutes, task = decide whether the school should ban phones in class and reach a group recommendation. Candidate B speaks roughly 70% of the turns, interrupts three times, and twice pulls the group into a tangent about social media and sleep. Candidate D speaks twice, briefly, but both contributions are clear, accurate and directly relevant - she proposes a compromise (phones in lockers during lessons) that the group eventually adopts. You must give each candidate an individual score for fluency, interaction and intelligibility. The other two candidates spoke occasionally and mostly agreed with B.
1. How do you score D's interaction when the task design gave her almost no room?
Read the worked answer
Separate opportunity from performance, and rate only performance - but record the constraint.
Her two turns contain a genuinely interactive move: she responds to the disagreement by proposing a compromise that the group then adopts. That is evidence of interaction, not just of speaking. So interaction is not zero; it is demonstrated, briefly, and at a reasonably high level.
What you must not do is infer a high score from two turns, or a low one from silence you caused. Where a trait needs sustained evidence and the sample is too thin, the honest score is a conservative one with a written reason: 'limited sample; task design suppressed output; score reflects observed evidence only.' Never let the thin sample contaminate fluency and intelligibility, which can be judged if her two turns were long enough to hear - hers were clear and accurate, so those traits should stand on their own evidence.
B is the mirror case: high fluency, but three interruptions and two tangents are interaction and discourse-management liabilities, not assets. Volume is not the trait being scored.
2. What can you do during the task without invalidating the ratings?
Read the worked answer
Any intervention changes the sample, so the default is: do not intervene mid-task. Run it as designed, then document the imbalance so the score is read in context.
If your protocol explicitly allows examiner moves, then the safest option is a procedural one applied to everyone, not a rescue aimed at D. For example, announce at minute six that each candidate must state a final position in turn. That is a rule change, applied uniformly, and it must be noted for all four candidates.
Nominating D directly ('D, what do you think?') is the move to avoid: it singles out one candidate, gives her a turn the others did not need help to get, and creates a rating you cannot compare. It also tells B that dominating works.
If the imbalance is severe enough that D's sample is unusable, the cleanest answer is a second round with a turn-forcing format, scored as a supplementary sample, rather than a contaminated first round you then have to explain.
3. Redesign the task so this does not recur.
Read the worked answer
The root problem is that the task has no information gap - nothing requires the quiet candidates to speak, so dominance is free.
Three fixes, in order of strength:
- Jigsaw: give each candidate one piece of information the group needs (a cost figure, a survey result, a parent complaint, a teacher's report). The recommendation cannot be reached unless all four contribute. This makes D's silence a group problem, not a personal trait.
- Assigned roles: chair, note-taker, devil's advocate, timekeeper - each with a required spoken output. Rotate roles across tasks so no candidate is always the note-taker.
- Required structure: shorten to six minutes; the first four minutes are round-robin (60 seconds each, no interruption), the last two are open. This guarantees every candidate a clean, un-interrupted sample for fluency and intelligibility before the discussion gets messy.
Keep one unstructured discussion somewhere in the test, because managing a dominant speaker is itself a real interaction skill - but do not let it be the only evidence you have.