How TeachProof Works

Safe to Use

TeachProof supplements clinical field experience. It does not replace it.

Simulation data is professional development data. It is not evaluation data, tenure data, or employment data. It cannot and should not be used for teacher evaluation or employment decisions. A template Union MOU is available for districts.

Grounded in Research

Every coaching suggestion in TeachProof cites published research. Coaching in TeachProof is informed by two widely used observation frameworks. It doesn't produce a Danielson or CLASS rating.

Danielson Framework

Charlotte Danielson (2013).

CLASS

Pianta, La Paro, & Hamre (2008). Classroom Assessment Scoring System.

Specific findings that drive the coaching engine:

Hattie (2009) — Feedback has an effect size of d=0.73, making it one of the highest-impact teaching strategies. TeachProof provides specific, immediate feedback on every teaching move.

VanLehn (2011) — A review of experiments comparing human tutoring, intelligent tutoring systems and no tutoring found an effect of d = 0.79 for human tutoring and d = 0.76 for step-based tutoring systems, well below the 2.0 often assumed. Tutoring research generally favors prompting a stuck student to reason over explaining for them (Chi et al., 2001), so TeachProof treats a question at an impasse as the stronger move, and an explanation offered before the student has tried as the weaker one.

Dweck (2006) — Person praise (“you're smart”) reduces persistence after failure. Process praise (“your strategy of...”) builds resilience. TeachProof automatically detects praise type and coaches accordingly.

Chi & Wylie (2014) — The ICAP framework: Interactive > Constructive > Active > Passive learning activities. TeachProof tracks ICAP level per student and flags when teaching moves reduce activation.

Rowe (1986) — 3+ seconds of wait time after asking a question increases response quality and participation equity. TeachProof scores “wait” as a valid teaching move, not inaction.

Rosenshine (2012) — Effective teachers check for understanding frequently during instruction. TeachProof flags long stretches of explanation without a comprehension check.

Live mentor observation with Watch, Coach, and Play modes. Post-session Socratic debrief powered by session-specific AI coaching. 20 archetype speech profiles ensuring each AI student responds with distinct, research-grounded behavior patterns.

K-12 Coaching

Hattie (2009), Rosenshine (2012), Lemov (2010), Dweck (2006), Black & Wiliam (1998), Kounin (1970). 155 K-12 scenarios with 19 student archetypes.

Higher Education

Ambrose et al. (2010), Freeman et al. (2014), Winkelmes TILT (2016), Kuh (2008), Finkelstein (2006). 12 faculty scenarios including AI integrity, seminar facilitation, and grade disputes.

Tutoring

VanLehn (2011), Chi et al. (2001), Wood & Wood (1996), Lepper & Woolverton (2002), Kraft & Falken (2021). 30 tutoring scenarios with context-aware coaching.

Context-aware coaching: K-12 sessions cite classroom research (Hattie, Rosenshine, Lemov). Higher education sessions cite faculty development research (Ambrose, Freeman, Winkelmes). Tutoring sessions cite tutoring-specific research (VanLehn, Chi, Wood). The coaching engine selects citations based on instructional context — not one-size-fits-all.

Honest About AI

TeachProof uses AI to power virtual student responses and generate coaching suggestions. We label everything clearly:

Measured — A core set of metrics taken directly from session data.

AI-estimated — Extended metrics labeled as AI estimates. Validation studies planned.

Context-aware — Coaching feedback adapts to the instructional context. A K-12 teacher, a college professor, and a private tutor receive different research citations appropriate to their practice.

TeachProof does not claim to replicate a real classroom. The goal is deliberate practice of specific teaching moves in a safe, repeatable environment.

How Practice Works

TeachProof is available on-demand. No scheduling, no booking, no live actors required.

Always available — Practice at any hour, from any browser. The simulator is always ready.

19 distinct students — Each AI student has a unique personality, speech pattern, emotional triggers, and internal thought process.

Mentor when you want one — Invite a supervisor to watch, coach, or take over a student avatar — or practice independently.

$0 per session — Free tier includes 5 sessions per day. No credit card, no trial expiration.

Data & Privacy

FERPA compliant. No student PII — all students are AI-generated. Teacher data owned by the teacher. Districts see only aggregate anonymized data. AI models do not train on teacher data.

Classroom simulation showing real-time state changes

Engine responsiveness — student states change based on teacher input

Research-grounded coaching in a high-stakes scenario

Try a Practice Session Contact Us