Attento AI · Research Brief

Improving Attention and Memory

How to structure online modules, multimedia, and asynchronous interactions so they work with — not against — the way attention, encoding, and long-term retention actually function.

Modeled attention across a 45-minute session

Toggle the design approach to see the difference structure makes.

At minute 10
At minute 25
At minute 45
Session average

Illustrative model, not raw data — the shape is informed by published findings on video engagement decay and the effect of segmenting and retrieval breaks on sustained attention. Guo, Kim & Rubin, 2014 · Mayer, 2021 · Roediger & Karpicke, 2006

01 Capture

Attention management

Working memory and sustained attention are limited, depleting resources. Long, unbroken stretches of instruction reliably outpace them — regardless of how interesting the material is.

Mechanism

Cognitive Load Theory Sweller, 1988 distinguishes three demands on working memory: intrinsic load (the material's inherent complexity), extraneous load (demands created by poor design — clutter, irrelevant detail, awkward navigation), and germane load (the effortful processing that actually builds understanding). Online modules that minimize extraneous load leave more working-memory capacity for germane processing.

Evidence — video length and engagement

An analysis of 6.9 million video-watching sessions across four edX MOOC courses found that video length was the single strongest predictor of engagement: watching time dropped sharply once videos passed roughly six minutes, and few students completed videos longer than nine. Guo, Kim & Rubin, 2014

6.9M
video sessions analyzed
~6 min
engagement inflection point
4
edX MOOC courses studied

Design moves: cap video and lecture segments around 6–9 minutes; open each module with a one-line roadmap that signals its structure; strip decorative content that doesn't carry meaning; let learners control pacing (pause, replay) rather than forcing a fixed runtime.

02 Process

Multimedia design

Richard Mayer's Cognitive Theory of Multimedia Learning Mayer, 2021 describes how learners build understanding from words and images through separate but limited-capacity channels. A handful of its principles do most of the practical work in course design:

Segmenting

Break a lesson into learner-paced chunks rather than one continuous stream — let people advance when ready.

Signaling

Use headings, highlighting, or a brief verbal cue to point attention at what matters most in a slide or video.

Coherence

Cut interesting-but-irrelevant detail, music, or imagery — it competes for attention rather than supporting it.

Spatial & temporal contiguity

Place related text and graphics near each other on the page, and narrate visuals as they appear, not before or after.

Modality

Narrate visuals aloud rather than stacking the same words as on-screen text — talking and reading compete for the same channel.

Pre-training

Introduce key terms and concepts before the main lesson so working memory isn't split between vocabulary and content.

03 Reinforce

Asynchronous interaction design

Retrieval — actively pulling information back out of memory — drives durable encoding more reliably than re-exposure to the same material. Discussion prompts, quizzes, and reflection tasks should be built as retrieval opportunities, not just engagement requirements.

Evidence — the testing effect

Across two experiments, students who took a recall test after reading a passage retained more on a delayed test (after two days or a week) than students who simply restudied the material the same number of times — even though restudying felt more effective immediately afterward. Roediger & Karpicke, 2006

Click to check off what your current course design already does:

0 of 4 design moves in place

04 Retain

Encoding & long-term retention

What learners remember weeks later depends less on how the material was first presented and more on how review and practice were distributed over time.

Spacing effect

A meta-analysis of 839 assessments across 317 experiments found that distributing review over time consistently produces better long-term retention than massing the same amount of study into one session. Cepeda et al., 2006

Interleaving

Mixing problem types during practice — instead of blocking one type at a time — improves long-term transfer, even though it feels slower and harder in the moment. Rohrer & Taylor, 2007

Why the harder version works

Spacing and interleaving are examples of what Robert Bjork termed desirable difficulties: conditions that slow performance during practice but strengthen retrieval pathways, producing better retention later. Bjork, 1994

Design moves: build weekly review touchpoints that reference earlier modules, not just the current one; mix topics within problem sets instead of grouping by type; replace a single end-of-term review session with several smaller spaced ones.