All design decisions

Measurements support teacher judgement

Why the app includes grade, language, and uncertainty in its measurements, and leaves the final judgement to teachers.

Teachers retain final judgement

Individual automated scores can differ substantially from a teacher’s.

Trade-off
Teachers need time to review recordings and measurements.

Use both word reading and comprehension for the suggested profile

Phil-IRI scores word reading as (passage words minus miscues) divided by passage words, and comprehension as correct answers divided by questions. Each becomes a percentage. Word reading of at least 97% is Independent, at least 90% is Instructional, and lower is Frustration. Comprehension of at least 80% is Independent, at least 59% is Instructional, and lower is Frustration. READSmart suggests the lower of the two levels, not a weighted average. For example, 98% word reading and 5 of 7 correct answers (71.43%) suggest Instructional, not Independent. WCPM measures speed separately and does not set the level.

Trade-off
The automated word score initially counts only omissions, substitutions and insertions. If answers are missing, the suggestion uses word reading alone; without a word score, it cannot suggest a level. The teacher listens, corrects the scores and chooses the final level for an official assessment.

Grade and language are required calibration features

Passage type and speech-recognition behavior change across grade and language.

Trade-off
A result cannot be calibrated without both fields.

Show measurements with uncertainty

Readers need the observed error ranges to interpret a score.

Trade-off
Reviewing a result takes more reading than checking a score alone.

Separate research findings from product choices

A product decision can involve classroom or privacy needs that an experiment did not test.

Trade-off
Readers may need to consult both sections to understand a choice.
Evidence
All findings