Skip to main content

Methodology

The science behind the design, with the design kept honest.

Every design decision here rests on published work. The citations are real and the claims about the product are limited to what the code does. There is no pilot data yet, so none is quoted.

1. Adaptive difficulty

Work in the zone where learning happens.

Vygotsky (1978) described the zone of proximal development: the gap between what a learner can do alone and what they can do with help. Tasks inside it, hard enough to need effort and reachable enough to avoid defeat, produce the most learning. One lesson aimed at the middle of a class misses the students on either side.

Elo (1978) built a rating system for chess in which a rating is a probability, not a fixed score: given two ratings, you can predict the chance of a result, and the ratings update by how far the result missed the prediction. The same idea works for a student and a question.

What Gregarious does

  • Each question carries a difficulty from 1 to 5 and each student a skill rating on the same scale, per standard.
  • The next item is chosen so the student should get roughly seven in ten right (a target near 0.72), hard enough to move and not so hard they shut down.
  • A right answer never lowers a skill rating and a miss never raises one. Automated tests hold that line.
  • Every third item tells the student “Adjusting for you,” so the adjustment is visible instead of hidden.

Vygotsky, L. S. (1978). Mind in Society. Elo, A. E. (1978). The Rating of Chessplayers, Past and Present. On desirable difficulty: Bjork, R. A. (1994), in Metacognition: Knowing about Knowing.

2. Mastery before creation

Why the reward is making something, not a badge.

Deci and Ryan's self-determination theory (1985; 2000) names three needs behind durable motivation: competence, autonomy and relatedness. Lepper, Greene and Nisbett (1973) showed the overjustification effect: an expected external reward for something a child already enjoyed reduced their later interest in it. Points and badges alone tend to run out.

Bloom's taxonomy (1956; revised by Anderson and Krathwohl, 2001) puts creating at the top of the cognitive ladder. Gregarious sequences the ladder instead of splitting it: practice questions ask for recall, comprehension and application; Studio asks for synthesis.

What Gregarious does

  • Studio opens when a student's saved sprint results reach 70% on a standard. Below that, the tools stay visible with “N points away.” Nothing is saved for a student until a parent or guardian opts in, so until then a sprint at 70% opens Studio on that device for the rest of the day.
  • Studio work is never graded. It is the student's: stories, comics, games, beats, pixel art, and writing shared with the class.
  • Relatedness is the engine, not a distraction: shared work, class galleries and class-level boards.
  • Whether creative time holds engagement better than points is a question the pilot will measure, not a result we have.

Deci, E. L., & Ryan, R. M. (1985). Intrinsic Motivation and Self-Determination in Human Behavior; (2000) Psychological Inquiry, 11(4). Lepper, M. R., Greene, D., & Nisbett, R. E. (1973). Journal of Personality and Social Psychology, 28(1). Anderson, L. W., & Krathwohl, D. R. (2001). A Taxonomy for Learning, Teaching, and Assessing.

3. Competition without a single loser board

Many boards, so almost everyone wins somewhere.

One leaderboard makes one winner and a room of losers, and the students at the bottom stop looking. The fix is not to remove competition, which motivates many students, but to rank more than one thing. Growth counts as much as level: a student who moves from 30% to 55% outranks one who stays flat at 85%.

What Gregarious does

  • Seven boards: Effort (EP this week), Growth (mastery gained), Studio (minutes this month), Games, Class vs Class (size-adjusted), Grade, and National, opt-in under a generated handle.
  • Streaks never go to zero. A missed day is “Starting a new streak.”
  • Hearts refill and a school can turn them off. Sprints have no time limit and award no points for speed, so a student who needs more time is never behind for taking it.

4. Teacher data

Evidence first, then the number.

A gap should be judged on all the evidence, not one quiz. Gregarious treats every sprint and every test a teacher logs as a data point, and gives each standard one conclusion: a big gap (several data points, mostly below 70%), a minor gap (below 70% on the latest work but right most of the time), an early signal, or closed. The data points are one click away. Attendance, behavior and discipline records are not used.

Growth Above Replacement (GAR) borrows baseball's question: given where these students started, how much more did they grow than expected? It would control for starting level, so a teacher who moves students two grades behind by a grade is not outranked by one whose students started ahead. It is in development, not live, and every page that mentions it says so. In development, Fall 2026

Every rule, with its numbers

Teachers see the same list in Settings. Each call is a suggestion; the teacher's call stands.

Studio

  • Studio opens at 70% on any standard, from saved sprints.
  • Nothing is saved for a student until a parent or guardian opts in, so until then a sprint at 70% opens Studio on that device for the rest of the day.
  • Studio work is never graded.

Sprints

  • Each next question is the one closest to the student’s level on the standard, preferring questions not seen in recent sprints.
  • After 3 right in a row, the next question reaches a little higher.
  • Between questions that fit the student’s level equally well, one whose reading is about something they picked on About me comes first. The fit to their level always decides first.
  • Students like you: of the students at the same school once within 10 points of the student’s score on a standard, how many reached 70% and the median days it took, said only when there are 5 or more.

Questions

  • What the answers say: a question is listed for a look once it has 10 answers and fewer than 30% are right, or one wrong answer draws 50% or more, or students doing well on the rest of the sprint miss it more.
  • Passages the answers flag: a passage with two or more questions is listed when students do 10 points or more worse on its questions than on the other questions of the same sprints, once it has 20 answers from 5 sprints. Each question shows its own percent, so a hard passage and one bad question look different.
  • Readings students liked: a passage counts once a sprint when a student taps “I liked this reading”, and is ranked on Item review once it was in 10 sprints. Counts only, never who; a like never changes which passages a sprint shows.
  • Difficulty from the answers: once a question has 20 answers and they put it a level or more from where it is set, Item review suggests the new level. A teacher decides; nothing moves on its own.
  • A question whose tutor hints students flagged 2 or more times is listed on Item review.
  • Read these first: up to 3 standards your classes are using, a sprint assigned now first, then the most sprints finished in the last 30 days.
  • Your share: the questions waiting are split across the teachers at your school by the grades their classes are, so each reads a part. A verdict from anyone counts, and anyone can read any question.

What followed

  • Your practice: each kind of move you made in the last 120 days, measured like “Did it work?”, said once it has 10 before-and-after scores. Moves under 10 points apart read as about the same. Only you see it.
  • School-wide, a kind of move is said only when 3 or more teachers made it, and no teacher is named.
  • Passages on their interests: percent right on readings about something students picked on About me against the rest, said once each side has 30 answers. Under 5 points apart reads as about the same.
  • What the app graded: your grading minutes in the weeks the app scored the most sprints against the rest, once 4 finished weeks have both. No minutes are estimated for the app.
  • Every line here says what followed, never what caused it.

Learning gaps

  • Big gap: 3 or more data points, at least two in three below 70%, and under 60% overall.
  • Minor gap: 3 or more data points and still open, but not low enough for a big gap.
  • Early signal: Only 1 or 2 data points below 70% so far. One more will confirm it or clear it.
  • Closed: The latest work, and the last 3 on average, are at 70% or better.
  • When a gap’s newest data point is over 30 days old, it says so: a quick sprint shows whether it is still a gap.
  • Your call on a gap stands until a newer data point arrives.

Almost there

  • Students at 55% to 69% on a standard that has a sprint, up to 3 standards, most students first.

After a warm-up

  • Under 15% missed: move on, or a follow-up with the students who missed it when 2 or more did and their sprints agree they are under 70%. 15% to 39%: a small group. 40% or more: reteach the class. Fewer than 5 answers: no call.
  • What your call did, for each student who missed it. Closed: the student reached 70% on the standard in a sprint within 14 days of the call.
  • Your calls are summed by kind, and by whether they matched the suggestion, once a group has 2 calls and 5 students with two weeks passed.

Since they started

  • Each standard’s first sprint against the latest, for the same students, once 3 or more have two sprints on it.

Short writes

  • A short write’s drafted score is written by AI, marked “AI draft: you decide,” and waits for your score; it is never the score until you confirm it.
  • The bias check compares how the drafts differ from your scores for students under 70% on the standard and for the rest, once each side has 3 or more answers you scored.

5. The teacher side

Measured with an instrument, not a testimonial.

Ertmer (1999) separated first-order barriers to technology use (access, time, training) from second-order ones (beliefs). The EdTech Fatigue instrument extends that line: 12 items on a 7-point scale, three subscales (Accumulated Skepticism, Emotional Exhaustion, Protective Withdrawal), approved by the IRB for a doctoral dissertation at Notre Dame of Maryland University. In the planned pilot, the five teachers would take it before and after the semester.

Ertmer, P. A. (1999). Addressing first- and second-order barriers to change. Educational Technology Research and Development, 47(4). Full design on the research page.

See the engine adjust.

Five questions, two minutes, and the third one tells you it is adjusting.