The Research Behind Your Education

In 1984, educational psychologist Benjamin Bloom published what would become one of the most cited findings in the history of education research. Working with graduate students at the University of Chicago, Bloom compared three instructional conditions: conventional classroom teaching (one teacher, 30 students), mastery learning in a classroom setting, and one-to-one tutoring combined with mastery learning. The results were striking. Students who received individual tutoring performed two standard deviations above the mean of conventionally taught students. In practical terms, this means the average tutored student outperformed approximately 98% of students in a traditional classroom.

Bloom called this the "2 Sigma Problem" -- not because the finding was problematic, but because the challenge it posed was enormous. If one-to-one tutoring could produce such dramatic improvements, how could educators achieve similar results at scale? Hiring a personal tutor for every student is financially impossible for any educational system. For four decades, this problem remained largely unsolved. Various approaches -- peer tutoring, smaller class sizes, improved curricula -- produced improvements but none approached the 2-sigma effect of individual tutoring.

St. Mary's's AI tutoring system represents the most promising approach to date for solving Bloom's 2-Sigma Problem. Our AI tutor provides individualized instruction that adapts to each student's pace, asks clarifying questions when comprehension falters, offers alternative explanations when a concept does not land, and provides immediate feedback on every response. While we do not claim our AI tutor perfectly replicates a master human tutor, the core mechanism Bloom identified -- personalized, responsive, one-to-one instruction with immediate feedback -- is precisely what our system delivers, available to every student at any hour of the day.

In their landmark 1990 meta-analysis, Kulik, Kulik, and Bangert-Drowns examined 108 controlled evaluations of mastery learning programs across multiple educational levels. The central premise of mastery learning is straightforward: students should demonstrate thorough understanding of one unit of material before moving to the next. The meta-analysis confirmed that mastery-based approaches consistently produced positive effects on student achievement, with stronger effects observed when programs were well-implemented and when corrective instruction was provided for students who did not initially demonstrate mastery.

St. Mary's's AI checkpoint system is a direct application of mastery learning principles. When students complete a lesson segment, they must pass a comprehension checkpoint before advancing to the next section. Students who do not initially demonstrate mastery receive targeted feedback from the AI tutor, can review the material, and attempt the checkpoint again. This is not a punitive gate but a supportive one -- the research shows that students who achieve mastery before progressing build stronger foundations and perform better on subsequent material. Our system automates what Kulik et al. identified as the critical components: clear mastery criteria, formative assessment, corrective feedback, and the opportunity for additional learning time.

Rapid Serial Visual Presentation (RSVP) is a reading technique in which text is displayed one word or short phrase at a time in a fixed position on the screen, at a controlled pace. Emerging research in cognitive psychology and educational accessibility has demonstrated that RSVP-style presentation offers significant benefits for readers with Attention Deficit Hyperactivity Disorder (ADHD). The forced-focus nature of RSVP eliminates the need for saccadic eye movements across a page, reduces the visual complexity that can overwhelm neurodivergent readers, and channels attention onto a single focal point. Studies have found that ADHD readers using RSVP-based systems show improved reading comprehension and report reduced cognitive fatigue compared to traditional text presentation.

St. Mary's's Accelerated reading mode is built on RSVP principles. Students who select this mode experience lesson content presented word-by-word or phrase-by-phrase at an adjustable pace, keeping their visual attention locked on a single point. For students with ADHD -- who make up an estimated 5-7% of the global adult population -- this is not a gimmick but a research-backed accommodation that can make the difference between struggling through a page and genuinely absorbing it. By offering RSVP alongside traditional reading and audio modes, St. Mary's ensures that neurodivergent learners have a pathway designed with their cognitive profile in mind.

Paul Black and Dylan Wiliam's 1998 review article "Assessment and Classroom Learning" is arguably the most influential paper ever published on the role of assessment in education. After reviewing over 250 studies, Black and Wiliam concluded that formative assessment -- assessment conducted during the learning process with the purpose of improving instruction and learning, rather than merely assigning grades -- produced effect sizes between 0.4 and 0.7 standard deviations. These gains were among the largest ever found for educational interventions. Critically, the researchers found that formative assessment particularly benefited lower-achieving students, helping to close achievement gaps.

St. Mary's's AI checkpoint system is formative assessment implemented in real time at a scale that Black and Wiliam could not have imagined in 1998. After each lesson segment, students encounter comprehension checks that assess understanding and provide immediate, specific feedback. These are not summative tests that assign a final grade -- they are formative instruments designed to identify what a student does and does not yet understand, then guide them toward mastery. The AI tutor can identify partial understanding, correct specific misconceptions, and offer tailored explanations, fulfilling every criterion that Black and Wiliam identified as characteristic of effective formative assessment: clear learning goals, evidence of where learners are relative to those goals, and actionable feedback to close the gap.

Richard Mayer's cognitive theory of multimedia learning, consolidated in his 2009 book and supported by hundreds of experimental studies, established a principle that seems intuitive but had never been rigorously demonstrated at such scale: people learn better from words and pictures together than from words alone. Mayer's framework goes far beyond this basic insight, articulating a dozen evidence-based principles for multimedia instructional design, including the coherence principle (exclude extraneous material), the signaling principle (highlight essential material), the segmenting principle (present lessons in learner-paced segments), and the modality principle (present words as spoken narration rather than on-screen text when accompanied by graphics).

St. Mary's's multi-modal lesson delivery system was designed with Mayer's principles as explicit design constraints. Our Interactive Video mode presents content through synchronized slideshows with audio narration, applying the modality principle. Lessons are segmented into focused parts with natural breakpoints, applying the segmenting principle. AI checkpoints between segments ensure that students process and integrate each segment before encountering new material, applying the temporal contiguity principle. Our Podcast mode provides audio with synchronized transcript highlighting, giving learners dual-channel processing. And our Read mode offers clean, focused text without extraneous visual clutter, applying the coherence principle. Mayer's research is not just cited in our design documents -- it is embedded in every screen, every lesson flow, and every interaction our students experience.