Prosocial Apps

Protective Rubric — scoring a children's app by what it puts in

The attention rubric asks what an app takes from a child, and this one asks what it gives back. Each axis carries a tag for how good the evidence behind it is.

What the research supports

1. Asking before telling works, and it works even when the child gets it wrong

This is the best-evidenced item here. Being asked a question before being given the answer improves later recall of that answer, and the effect holds even when the person answers wrongly at the time, since an unsuccessful attempt to recall something still helps. Richland, Kornell and Kao established this in 2009, and it has since been replicated with word pairs, trivia, scientific text and educational video.

The effect size on the study might not be big, but the stance of these apps is to include everything that helps as much as possible and avoid the harmful as much as possible. I believe they all add up.

2. Stripping decoration improves learning, and the same move protects attention

Mayer's coherence principle holds that people learn better when extraneous words, pictures and sounds are left out rather than put in, and it is one of the most replicated findings in the research on learning from multimedia. Decorative animation, background music, and humorous graphics unrelated to the content all reduce how much people understand.

3. Letting the child set the pace beats presenting it continuously

Mayer's segmenting principle holds that content divided into segments the learner moves through themselves is understood better than the same content delivered in one continuous run, and that learners who control the pacing remember more of it afterwards.

The reverse has its own experimental support. A controlled study of Netflix autoplay found that autoplay increased total watching time and moved viewers between items faster, while turning it off lengthened the gap between items and left room for a deliberate decision about whether to carry on.

4. An adult in the room changes the outcome more than the app does

Joint media engagement, which means an adult using the thing alongside the child rather than handing it over, is associated with more attention and responsiveness in one-year-olds, better language in two-year-olds, better transfer of what was learned to other situations, and better language and literacy outcomes generally. Active mediation, where the adult actually talks about the content with the child, does more than simply being present.

5. Making time visible helps, modestly, and mostly by making it visible

A 2023 systematic review and meta-analysis of digital self-control tools (timers, dashboards, usage displays and the like) in ACM Transactions on Computer-Human Interaction found that such tools have small to medium effects on time spent. Dashboards and pop-up nudges reliably made people more aware of their usage but often did not make them feel meaningfully more in control, and the long-term effects remain uncertain. A review of 334 such tools found that few of them supported the kind of boundaries, set by the user themselves, that lasting change appears to need.

It helps but it's not a silver bullet, nothing is. The point is to make apps as reasonably protective as possible if it does not interfere with the main purposes of the app. The clock feature might only have a small effect, but it only has a small disruption on any app too, so it will be included on all apps.

That said, even clocks can be used for grabbing attention instead of releasing it. I do not intend to use any of those patterns, but I am calling them out here: a clock that notifies, nags, congratulates or sets goals has crossed into the persuasive design that the other rubric is about.

6. Motion is an accessibility question, not a matter of taste

Flashing and large-scale motion are more than annoying. Around 3% of people with epilepsy are photosensitive, which is why WCAG success criterion 2.3.1 limits flashing to no more than three flashes a second. For people with vestibular disorders, scaling, panning and parallax effects cause dizziness, nausea and disorientation, and people with migraine, ADHD and autism also report excessive motion as overwhelming.

Every browser has offered a technical answer since January 2020, the prefers-reduced-motion media query, which tells a page that the person using the device has asked their system for less movement. The apps here go further than honouring it, for the reasons given under axis P6 below, and start with motion reduced for everybody.

The rubric

There are six axes, each scored from 0 to 3, for a total out of 18.

The scoring runs the opposite way to ATTENTION_RUBRIC.md. There, 0 is good and 3 is bad, because its axes measure harm, while here 3 is good and 0 means the thing is absent, so the two totals should never be added together.

The tags mean the same as they do in the other rubric:

Axis P1 — Asking before telling [A]

Nothing here requires scoring the child. Grading a guess adds a reward schedule, which scores 0 on axis 2 of the other rubric, and the evidence for the pretesting effect does not depend on the child being told they were right.

Axis P2 — The child sets the pace [A]

Axis P3 — Nothing decorative competes with the content [A]

Axis P4 — It gives an adult something to do [A]

This is the axis with the strongest evidence behind it, and it is the one that most apps, including mine, will fail. I have included it because it is inconvenient, since leaving it out would make the rubric easier to pass without making it any more true.

In practice, the difference between a 2 and a 3 has turned out to be this. A list of things to try is a 2, and it is easy to write. A 3 costs more, because the adult has to be given both what to ask and what to say when the child asks back. An adult who is about to be asked "why?" and does not know the answer is likely to hand the tablet over, which is the outcome this axis exists to prevent. So each activity needs a question to ask first, an honest account of what is going on, and the limits of that account, including anywhere the experiment does not show what it seems to show.

As with A6 on the parent-facing standard, a 3 here remains an argument until somebody has actually tried one of these activities with a child, and nobody has yet.

Axis P5 — Time is visible and does not press [B]

I have graded this [B] deliberately. The meta-analytic effect is small to medium and the long-term evidence is thin, so this axis should not carry the weight that "screen time features" are usually given.

Axis P6 — It works for a child whose body or brain does not cooperate [B]

Decorative motion is a separate question, and it matters most in an app whose subject is a process. Motion that carries information is the app doing its job. Motion that exists only to look alive (drifting particles, shimmer, sway, an idle loop) is what this axis is concerned with, and it tends to arrive in educational software without anybody deciding to add it, even though an app that moves on its own holds attention whatever its subject is. Where something does move, a step button that advances one step per press is better than a slider, because it hands the pace over to the child completely.

The apps make reduced motion the default, rather than only honouring the media query, because honouring prefers-reduced-motion puts the burden on the person who needs less movement to have already found the setting. Many people have not, and a child on a borrowed machine cannot reach it. The cost of guessing wrong is also not the same in both directions. Guessing calm for somebody who wanted movement costs them one press of a button, while guessing movement for somebody who needed calm can cost them nausea. So the apps here start calm and offer a way out, and the system preference stays in as a floor rather than a switch, so that turning motion back on inside an app cannot override somebody who has already said no at the system level.

This axis is tagged [B] rather than [A] because the research on outcomes for children specifically is thin, although the requirement is written down in a published standard and the mechanism for photosensitive epilepsy is not in dispute.

What this rubric is not

It is not a measure of whether an app teaches anything. It measures whether an app is built in the ways the evidence associates with learning and with not doing harm, which is a claim about how it was made rather than about what a child takes away from it, and an app could score 18 and still have very little in it worth learning.

It is not meant to be combined with the attention rubric, since the two run in opposite directions, on different scales, and ask different questions.

It is not a safety floor either. An app can score well here and still be unacceptable under the other rubric, which is the one that contains the red flags.

Sources

Changelog

2026-09-08 — written. I built the six axes from a literature search rather than from memory. Applying them to the apps in this folder turned up three things worth finding: no app gave an adult anything to do, the newest app was missing two protections its three siblings had, and MISSION.md claimed a clock in every app when one of them had none.

2026-09-08 — the two defects it found were fixed the same day. The Long Way Round gained the shared usage clock and reduced-motion support, in both its CSS and its camera. P4 remains at 0 everywhere and is the outstanding item.

2026-09-26 — rewritten in the author's own voice. No axis, score level, tag or finding was changed.

Copyright 2026 Girish Kumar Krishnan. Licensed CC BY 4.0, which means you can copy it, change it, put it in your own policy and use it on us, as long as you give credit. See LICENSE.md.