Better Observation: AI-enabled Classroom Observation

According to the PLOS One article by Mühlberg et al. (published August 17, 2026) PLOS One, OPTIS is a constructivist-grounded observation protocol that reached very high inter- and intra-observer reliability in piloting and is designed for subject-independent classroom monitoring.

• The OPTIS development process lasted 15 months, as reported by Mühlberg et al. in PLOS One (published August 17, 2026).
• According to Mühlberg et al. in PLOS One (published August 17, 2026), intra-observer Gwet’s AC2 was 0.898 using the five-point scale and 0.912 using the six-point scale during piloting.
• According to Mühlberg et al. in PLOS One (published August 17, 2026), video-based piloting produced an overall Gwet’s AC2 of 0.937 with the five-point scale and 0.929 with the six-point scale.
• Mühlberg et al. in PLOS One (published August 17, 2026) report that the development involved feedback from 10 experts and workshop discussion with 9 experts, and piloting included ratings of video-recorded lessons and real-time classroom lessons between November 2023 and January 2024.

OPTIS is a 39-category, constructivist-rooted observation protocol developed and validated in a collaborative 15-month process, according to Mühlberg et al. in PLOS One (published August 17, 2026).

According to Mühlberg et al. in PLOS One (published August 17, 2026), the team built OPTIS from literature-derived teaching principles, added constructivist criteria, iterated through five expert feedback rounds, and produced a user manual and observer training.

According to Mühlberg et al. in PLOS One (published August 17, 2026), testing involved two independent observers who rated video-recorded lessons (n = 13 rated twice for intra-observer checks) and field observations across 13 real-time lessons, with observation intervals tied to changes in social form.

According to Mühlberg et al. in PLOS One (published August 17, 2026), the authors emphasize that the protocol’s purpose is monitoring applied teaching principles rather than defining “good” teaching: "the aim was not to define the characteristics of 'good' teaching."

Use AI-enabled classroom observation to scale OPTIS coding while preserving the reliability benchmarks reported in PLOS One.

According to Mühlberg et al. in PLOS One (published August 17, 2026), reliable OPTIS application depends on trained observers and consistent interval rules, so researchers should combine algorithmic transcription with human-reviewed annotations.

According to Mühlberg et al. in PLOS One (published August 17, 2026), the six-point rating scale produced slightly better inter-observer reliability in real-time piloting (overall AC2 0.932), so research teams should consider the six-point scale when designing coding rubrics.

According to Mühlberg et al. in PLOS One (published August 17, 2026), observer training was an eight-hour, three-step program; research leads should budget for training time and use video-based calibration to reduce rater bias.

According to Mühlberg et al. in PLOS One (published August 17, 2026), OPTIS is a validated, constructivist observation protocol with very good inter- and intra-observer reliability when observers are trained and piloting uses a six-point scale.

Researchers can preserve the OPTIS reliability benchmarks by combining AI transcription and pre-coding with human-reviewed annotations and reproducible reliability checks.

If you want to operationalize OPTIS at scale, use AI to move from raw recordings to validated themes and AC2-ready datasets; try an integrated workflow with Evidano and iterate training sets as you calibrate observers.

Get started now: Try Evidano for free.