The Studyh method

Active recall, also called retrieval practice, means trying to produce information before consulting the source. In a study using prose materials, taking memory tests produced more delayed retention than repeated study, although repeated study raised confidence on the immediate test. ^[1]

Karpicke and Blunt also compared retrieval practice with elaborative concept mapping. In that experiment, retrieval produced more learning on comprehension and inference questions. That is evidence for that experimental design, not a rule that retrieval is always better in every situation. ^[2]

Distributed practice spreads study opportunities over time instead of concentrating them in one session. Cepeda and colleagues found that the relationship between the gap between sessions and the final retention interval depends on the conditions studied. The review therefore does not establish one ideal interval for every person, subject, or exam. ^[3]

SM-2 belongs to the history of systems that tried to turn performance into review intervals. Woźniak and Gorzelańczyk published a 1994 paper on optimizing repetition spacing; it helps place the method historically, but does not automatically validate a schedule for every learner. ^[5]

Recognition means identifying an option or feeling familiarity when information appears; retrieval means producing an answer with less external support. They are not the same task. Rawson and Zamary studied why testing effects can be stronger when practice involves free recall rather than recognition. ^[7]

In Studyh's deep mode, a session starts with the student's response before consulting the original block. A response evaluated well in one session still does not establish transfer to every situation or prove broad mastery.

The platform compares the response with the original block and key concepts to produce an operational indicator for that attempt. The score describes that response; it is not an objective measurement, certification, or proof of mastery.

A session average is a simple average of the scores included in that session. It summarizes those responses and is not independent evidence of material quality, question difficulty, or the ability to apply the content in a new situation.

Studyh uses a scheduling heuristic inspired by SM-2. Self-reported performance changes the ease factor, repetitions, interval, and next review date. The schedule is based on performance; it does not claim to predict an individual's exact forgetting or find a universally ideal interval.

The schedule is a product heuristic that reduces the need to decide manually when to return to an item. Evidence about distributed practice supports using spacing, but does not turn any particular schedule into a universal solution. ^[3]

An AI evaluation can be wrong or vary with the wording and context of an answer, and it can reproduce bias. Research on LLM-as-a-judge discusses position, verbosity, and self-enhancement biases, among other limitations. Agreement with human preferences is an approximation, not a certification of correctness. ^[6]

Treat feedback as a signal to investigate. Check the original material and, when the subject has high consequences — such as a professional rule, academic guidance, or official requirement — confirm it with an authoritative source and a qualified teacher or specialist.

Studyh helps make an attempt visible, but the responsibility for checking the source, interpreting context, and deciding whether an answer is sufficient remains with the student.