SL.1Part of it
Collaborative discussion
- Hear
- In a debate, whether you answered the argument that was made.
- Don't
- Turn-taking in a live class. A solo rep has nobody else in the room.
Standards
What we listen for
Every skill in the Oracy Framework, and whether we can hear it in a recording.
01Coverage
The four strands
Pick one.
Physical: 4 of 6 heard
- [heard] Pace of speaking
- [heard] Tonal variation
- [heard] Clarity of pronunciation
- [heard] Voice projection
- [not heard] Gesture and posture
- [not heard] Facial expression and eye contact
Linguistic: 3 of 3 heard
- [heard] Appropriate vocabulary choice
- [heard] Register
- [heard] Rhetorical technique
Cognitive: 5 of 5 heard
- [heard] Choice of content
- [heard] Structure and organisation
- [heard] Clarifying and summarising
- [heard] Self-regulation
- [heard] Reasoning
Social & Emotional: 2 of 4 heard
- [not heard] Working with others
- [not heard] Listening and responding
- [heard] Confidence in speaking
- [heard] Audience awareness
Framework
Oracy Skills Framework, Voice 21 in partnership with Oracy Cambridge (Hughes Hall Centre for Effective Spoken Communication). Used here as formative vocabulary, not as an assessment scale.
The line
We never judge these
None of it reaches your feedback, as praise or as a fault.
Why this comes first
Speech AI usually marks down AAVE, Spanish-influenced English and multilingual students. We built Vocius for those students, so we only check whether you were understood.
The one thing we flag
A word nobody could make out because it was mumbled or cut off. We name the word and when it happened. We never flag an accent.
It lives in the code
That list is read from the same file our prompts use. A test fails the build if a prompt ever brings one of them back.
02Assessment
There is no score
A rep never gets a grade, a mark out of ten or a rank. There is no leaderboard.
Your rep · 0:41Pace
and that's why I think the rule should change . You sped up here, about 30 words a minute faster than the rest.
Try next: say that line again, and put a full stop before “because”.
- One moment per rep, with the timestamp and your own words.
- If we couldn't measure something, it says unavailable, not zero.
- No trend line until there are enough reps to draw one.
The framework's authors say it was not built for assessment. We use it as vocabulary, the way they ask.
03Common Core
Speaking and Listening, mapped
The six anchor standards. One of them we can't touch at all.
SL.2Part of it
Integrating information
- Hear
- Whether the sources you name are used to back the claim.
- Don't
- Whether you understood a chart or a video. We only hear what you say about it.
SL.3Part of it
Evaluating reasoning
- Hear
- Whether you answered the reason behind a claim, or just the claim.
- Don't
- What you think while you listen. We can't hear that.
SL.4Most of it
Presenting clearly
- Hear
- Structure from the transcript. Pace, projection and tone from the audio.
- Don't
- Whether the room followed you. We only hear the speaker.
SL.5None of it
Media and displays
- Hear
- Nothing. Vocius listens to audio.
- Don't
- Slides, posters, props or handouts.
SL.6Part of it
Adapting to context
- Hear
- Whether your words fit the situation the prompt described.
- Don't
- “Command of formal English” as correctness. See the block above.
Sub-standards differ by state, so this maps the six anchors. Sit it next to your department's wording.
04Context
Why this hasn't been measured
Speaking is in the standards. Neither PARCC nor Smarter Balanced tests it. Both planned to, then dropped it when the tests were shortened.
A teacher can run a round of presentations once or twice a year. Vocius is for the reps in between.
150min
Thirty students, three minutes each, plus setup. That is most of a week of class.