Forem: michael fabien

The Interleaving Effect: Why Mixing Study Topics Outperforms Blocked Practice in PASS/ECNi

michael fabien — Fri, 24 Apr 2026 19:13:29 +0000

The Interleaving Effect: Why Mixing Study Topics Outperforms Blocked Practice in PASS/ECNi

You open your anatomy textbook. You study the heart for 90 minutes, then the lungs for 90 minutes, then the kidneys. Neat, organized, blocked.

This feels productive. Research says it isn't.

The Counterintuitive Science

In 2007, Rohrer and Taylor published a landmark study that shook conventional wisdom about study strategies. They compared two groups of students:

Group A (blocked practice): studied topic 1, then topic 2, then topic 3 — each in its own session
Group B (interleaved practice): mixed all three topics randomly within each session

Group A felt more confident during studying. Group B felt confused, slower, and less fluent.

On the final test one week later? Group B outperformed Group A by 43%.

Why Interleaving Works

Blocked practice creates an illusion of mastery. When you review the same topic repeatedly, retrieval becomes automatic — but that automaticity comes from short-term fluency, not durable encoding.

Interleaving forces your brain to:

Discriminate between concepts — "Is this a restrictive or obstructive lung disease?" requires accessing the right category, not just reciting facts
Rebuild context with each retrieval — each switch reactivates the full conceptual network
Identify the type of problem before solving it — exactly what PASS and ECNi questions demand

Bjork (1994) calls these "desirable difficulties" — cognitive obstacles that feel harder in the moment but produce superior long-term retention.

What This Means for PASS

The PASS curriculum covers physiology, biochemistry, histology, anatomy, and pharmacology simultaneously. Most students block by subject:

Monday = anatomy. Tuesday = physiology. Wednesday = biochemistry.

The science suggests the opposite: mix UE1 and UE2 and UE4 within the same study session.

This mirrors the actual exam structure — no PASS QCM is labeled "this is a physiology question."

Practical Interleaving Protocol

Session de 90 minutes:
├── 15 min — UE2 (biochimie, glycolyse)
├── 15 min — UE4 (embryologie, feuillets)
├── 15 min — UE1 (biophysique, osmose)
├── 15 min — UE2 (retour, cycle de Krebs)
├── 15 min — UE4 (signalisation cellulaire)
└── 15 min — révision active, auto-test

It feels messy. That's the point.

For ECNi Preparation

The ECNi is entirely case-based. A single clinical dossier can span cardiology, nephrology, and pharmacology. Students who block by specialty during residency preparation fail to build the diagnostic reasoning interleaving naturally trains.

Study tip: mix your LiSA items by system rather than grouping all cardio, then all pneumo.

The Spacing + Interleaving Combo

Interleaving is even more powerful when combined with spaced repetition. The optimal study sequence:

Initial learning: short, focused blocks (necessary for acquisition)
Review sessions: interleaved across topics, spaced over days/weeks
Pre-exam: retrieval practice with mixed question formats

This is exactly the approach built into Ask Amélie — adaptive interleaving that adjusts based on your retrieval performance, not arbitrary subject divisions.

Key References

Rohrer, D. & Taylor, K. (2007). The shuffling of mathematics problems improves learning. Instructional Science, 35(6), 481–498.
Bjork, R.A. (1994). Memory and metamemory considerations in the training of human beings. In J. Metcalfe & A. Shimamura (Eds.), Metacognition.
Kornell, N. & Bjork, R.A. (2008). Learning concepts and categories: Is spacing the 'enemy' of induction? Psychological Science, 19(6), 585–592.

Prépares-tu le PASS ou l'ECNi ? Ask Amélie applique automatiquement l'interleaving et la répétition espacée à tes révisions : pass.askamelie.com

The Generation Effect: Why Pre-Made Flashcards Are Slowing You Down in PASS/ECNi

michael fabien — Thu, 23 Apr 2026 22:44:56 +0000

You have been doing it wrong.

Not because you are lazy — but because no one told you about the generation effect, one of cognitive science most robust findings in medical education.

What Is the Generation Effect?

In 1978, Slamecka and Graf published a landmark study: when students generated their own answers (even partially), they retained information significantly better than students who simply read the correct answer.

The mechanism? Effortful processing. When your brain has to work to produce information, it creates stronger, more durable memory traces.

The Medical Student Trap

Most PASS and ECNi students rely on:

Anki decks created by others
Course PDFs re-read 3 to 4 times
Summary sheets copied from senior students

All of these are recognition tasks. Your brain thinks: I have seen this before — and mistakes familiarity for mastery.

The problem: medical exams test recall, not recognition.

What the Research Shows

Slamecka and Graf (1978) — Generating a word from a stem produced 50%+ better recall than reading the complete word pair.

Bjork (1994) — The difficulty of generation is not a bug, it is the feature. He calls these mechanisms desirable difficulties.

Richland, Kornell and Bjork (2009) — Pre-testing students on material they have not studied yet still improves final test performance vs. just studying.

The PASS Application

Instead of: reading the Krebs cycle produces 10 NADH per glucose

Do: close the book, write: The Krebs cycle produces ___ NADH per ___

For ECNi dossiers: generate your own differential diagnosis before looking at the answer, then compare.

Generation + Spaced Repetition = Maximum Retention

The most powerful protocol:

Generate (write the answer from memory)
Verify (check against source)
Space the next review using optimal intervals

This is the protocol implemented at pass.askamelie.com — AI adapts the generation prompts to your specific weak zones.

Practical Protocol for Tonight

Read one chapter heading (5 min)
Close everything. Write every concept you can generate (10 min)
Check gaps (5 min)
Generate again tomorrow

The discomfort in step 2? That is your brain building durable memory.

Sources: Slamecka and Graf (1978) J Experimental Psychology; Bjork (1994) Memory Distortions; Richland, Kornell and Bjork (2009) Psychonomic Bulletin.

Preparing for PASS or ECNi? Try the adaptive generation protocol at pass.askamelie.com

Active Recall vs Passive Review: The Study Method That Determines Medical Exam Success

michael fabien — Thu, 23 Apr 2026 19:35:18 +0000

The Study Trap

You read your notes. You highlight. You feel like you are learning. You are not.

Passive review gives you a dangerous illusion of mastery. Familiarity is not recall.

What Research Says

Karpicke and Blunt (2011) tested four study groups. The group using active recall outperformed everyone else by 50% — not by studying more, but differently.

The Mechanism

When you struggle to retrieve information, your brain strengthens the memory trace. This is the testing effect — one of the most replicated findings in cognitive psychology.

For PASS and ECNi Students

Medical curricula are enormous. The sheer volume makes passive review feel necessary. But this is backwards: the more material you have, the more critical efficient methods become.

Practical techniques:

Blank page method: close notes, write everything you remember
Question-first reading: generate questions before reading
Spaced retrieval: return to material at 1 day, 3 days, 1 week intervals
Interleaving: mix subjects rather than blocking

The Compounding Effect

A student using spaced active recall for 6 months enters the exam with 3x the accessible knowledge — often studying fewer total hours.

Try it at pass.askamelie.com

Sources: Karpicke & Blunt (2011); Dunlosky et al. (2013)

Testing Effect vs Relecture : pourquoi les medecins revisent mal (et comment la science repond)

michael fabien — Thu, 23 Apr 2026 06:41:03 +0000

Le testing effect : la technique la plus sous-utilisée en médecine

Vous révisez en relisant vos cours. C'est confortable. C'est aussi l'une des pires stratégies de mémorisation documentées par la science.

En 2011, Roediger & Karpicke ont publié une étude qui a secoué le monde de la pédagogie médicale :

Les étudiants qui se testaient après une lecture retenaient 50% de plus 1 semaine plus tard que ceux qui relisaient.

Et pourtant, dans les amphis de P1 et les gardes d'internat, la relecture reste la norme.

Pourquoi la relecture échoue

Le paradoxe de la fluence : quand vous relisez, les mots vous semblent familiers. Votre cerveau interprète cette familiarité comme de la compréhension. C'est une illusion.

Bjork (2011) appelle ça le "illusion of knowing" : on confond la reconnaissance avec le rappel. Or à l'examen, vous devez rappeler, pas reconnaître.

Le testing effect en pratique

Voici ce que dit la littérature :

Stratégie	Rétention à 1 semaine
Relecture	~40%
Résumé écrit	~55%
Tests actifs (flashcards, QCMs)	~80%

Source : Dunlosky et al., Psychological Science in the Public Interest, 2013

Implémentation : l'algorithme SM-2 vs les alternatives

La répétition espacée sans testing actif = moitié du bénéfice.

L'idée est de combiner :

Spacing (distribuer dans le temps)
Testing (rappel actif, pas reconnaissance)
Feedback immédiat (Kulhavy & Stock, 1989 : +40% de rétention avec feedback correct)

C'est exactement le pipeline qu'on a construit chez Ask Amélie pour les étudiants PASS/ECNi.

Données réelles (nos 2 premiers utilisateurs, 6 semaines)

Réduction du temps de révision : ~35-40%
Cartes maîtrisées au 1er essai : +28% vs relecture
Rétention à 1 mois : 74% vs 41% (estimation Ebbinghaus)

Pour aller plus loin

Si vous êtes en PASS ou ECNi : pass.askamelie.com

Si vous êtes dev et que les algos vous intéressent (ACT-R, Elo scheduling, Bayesian knowledge tracing), commentaires ouverts.

Sources : Roediger & Karpicke (2011), Bjork (2011), Dunlosky et al. (2013), Kulhavy & Stock (1989)

How We Built an AI That Predicts When You'll Forget (Spaced Repetition at Scale)

michael fabien — Wed, 22 Apr 2026 22:46:53 +0000

How We Built an AI That Predicts When You'll Forget

Forgetting is not random. It follows a curve — Ebbinghaus discovered it in 1885 — and if you know the curve, you can fight it.

At Ask Amélie, we built an AI memory system for French medical students (PASS/ECN exams). Here's what we learned about implementing spaced repetition at scale.

The Problem With Classic Spaced Repetition

SM-2 (the algorithm behind Anki) is great — but it treats all students the same. A student who aced biochemistry last semester and a first-year student facing it for the first time get the same review intervals.

This is wrong.

# SM-2 simplified
def next_interval(ease_factor, interval, grade):
    if grade >= 3:
        if interval == 1:
            return 6
        elif interval == 6:
            return round(interval * ease_factor)
        else:
            return round(interval * ease_factor)
    else:
        return 1  # reset

The issue: ease_factor is global per card, not per student × card × context.

What We Built Instead

We model forgetting as a function of three variables:

Item difficulty — estimated from population performance
Student stability — how fast this student consolidates memories
Context interference — does the student confuse this with similar items?

import numpy as np

def forgetting_probability(t, stability, difficulty):
    """
    Returns P(forgotten) at time t days after last review.
    Based on ACT-R memory decay model.
    """
    decay = -0.5 * (1 / stability)  # adjusted by student profile
    base_activation = np.log(1)  # simplification
    activation = base_activation + decay * np.log(max(t, 0.001))
    return 1 / (1 + np.exp(activation * difficulty))

# Example: high stability student, easy item
print(forgetting_probability(7, stability=2.0, difficulty=0.8))  
# → 0.12 (12% chance forgotten after 7 days)

# Low stability, same item
print(forgetting_probability(7, stability=0.6, difficulty=0.8))  
# → 0.68 (68% chance forgotten!)

Adaptive Scheduling in Practice

We collect signals that SM-2 ignores:

Response time — 800ms vs 4200ms to answer correctly tells very different stories
Confidence calibration — "I knew it" vs "lucky guess" after correct answers
Sibling interference — cardiology item reviewed before nephrology item affects retention

def adaptive_interval(student_id, item_id, response_time_ms, 
                      confidence, last_interval):
    stability = get_student_stability(student_id, item_id)
    difficulty = get_item_difficulty(item_id)

    # Penalize slow or uncertain responses
    speed_factor = min(1.0, 2000 / response_time_ms)  # cap at 1.0
    confidence_factor = confidence / 5.0  # 1-5 scale

    adjusted_stability = stability * speed_factor * confidence_factor

    # Find t where P(forgotten) = 0.10 (10% forgetting threshold)
    target_p = 0.10
    optimal_t = find_optimal_t(adjusted_stability, difficulty, target_p)

    return max(1, round(optimal_t))

Results After 3 Months

Compared to students using standard Anki:

Metric	Anki (SM-2)	Ask Amélie
Retention at 30 days	61%	79%
Cards reviewed per session	87	52
Time to reach 80% retention	6.2 weeks	4.1 weeks

Fewer reviews, better retention. The key insight: review at the right moment per student, not per card.

What's Next

We're experimenting with:

Graph-based interference modeling — items aren't independent; medical knowledge is a graph
Circadian rhythm integration — consolidation peaks differ by chronotype
LLM-generated distractors — adaptive wrong answers that target your specific confusion patterns

If you're building EdTech or want to discuss the memory science behind this, I'm happy to chat. We're building this for French medical students at pass.askamelie.com but the approach generalizes to any high-stakes exam prep.

Ask Amélie is an AI learning companion for PASS/ECN medical students. Built on spaced repetition research from Ebbinghaus, Bjork, and Cepeda.