PRACTICES & INTERVENTIONS
For You
FOR YOUR TEAMS
Structures, not suggestions. Everything in this section should be a default, a calendar rule or an agenda line. Things that work on people who will never attend a wellbeing session. The evidence for this exists: across roughly ninety individual-level wellbeing offerings in 233 UK organisations, almost none showed a measurable benefit to the people who took them up (Fleming 2024).
The one exception was scheduled, social, and part of the working day.
2. Meeting Time Defaults
Default meeting lengths to 25 and 50 minutes,
Set the calendar system's default meeting lengths to 25 and 50 minutes, organisation-wide.
3. Response Time Norms
Email is a four-hour medium
Chat should not an interrupt channel for non-urgent items.
T4. Decision Point Pauses on Critical Actions
Friction design where it matters
Add specific warnings and procedural check at key points (payments, credential entry, external data sharing).
T5. Quick Check and Feedback
Build how would I know if this were false or safe?
Make it easy in the workflow for people to ask, verify or double check if they have doubts.
T6. Feedback in the Moment
Without any shame
Where simulations are used, feedback delivered at the moment of failure cut re-victimisation two weeks later from 50% to 40% in a preregistered trial (Bender 2026).
T7. Review Time for AI Output
Reviewing AI output is real cognitive work and is currently invisible in every workload model. Make AI review time a named, budgeted activity;
07. Breath
Four in, four hold, four out, four hold
Two or three rounds of box breathing before something that matters.
Attention:
The Pause Before Reacting
What the evidence says.
Manipulation works by producing a reaction that arrives before reflection can. Alarm and account-compromise framings roughly triple phishing failure rates (Madeira & Alves 2026); uncertainty draws clicks (Tian & Jensen 2019); time pressure collapses the ability to tell real from fake without making people more or less trusting overall (Jones 2019; Butavicius 2022; Sultan 2022). And in a study of nearly a thousand people, feeling *any* emotion in response to a headline (pleasant or unpleasant) went with believing it more, true or false (Bago et al. 2022).
So the reaction is real information. What you do with it is the question. The popular answer: "calm down first" was tested by the same team in three preregistered experiments, 6,702 people: asking people to suppress or reappraise their emotions did nothing for discernment.
What did work however is redirecting attention to the message: asking how you'd know whether it's true reduced sharing of false content without touching true content (Fazio 2020), and attention applied to the specific message is the strongest protective factor in the phishing literature (Bera & Kim 2025).
How to do it.
-
Notice that you've reacted. Urgency, alarm, flattery, the small thrill of good news. Name it lightly, once. Don't work on it.
-
Step back half a pace. *There is urgency here* rather than *I must hurry*. The feeling is an event, not an instruction.
-
Turn toward the message *What does this want me to do and why now?* Then: *how would I know if it were false?*
-
Act on the second reading.* For anything irreversible (money, credentials, data) verify through a channel the message didn't hand you. Make that a rule, not a practice.
Caution.
Where this gets complicated: Steps one and two are the part of mindfulness with the best general evidence; step three is the part with the best evidence in this domain. The claim that noticing an emotion leads people to *read it as a signal about the message* is the link between them, and nobody has tested it here yet, that's my research.
Two further cautions: a brief pause before a persuasive message can deepen processing and so *amplify* a well-made one (Kupor & Tormala 2015), and one study found metacognitive awareness improved judgement of truth while increasing willingness to share false content (Pérez Sánchez 2024). Pausing by itself is not enough of an armour, but it provides the condition under which judgement can work at all.
"Almost every successful manipulation, a phishing email, an outrage post, a dark pattern depends on the same thing: a reaction that happens before reflection can. My research asks whether we can train that missing pause."
Anna Collard, Founder, Digital Mindfulness
How Would I Know?
One Question Before you Share
What the evidence says.
This is the best-evidenced five seconds in the whole catalogue so far. Fazio (2020, N = 501) asked people to explain how they knew a headline was true or false before deciding whether to share. Willingness to share at least one false story fell from 57% to 39%; sharing of true stories didn't move. The wider accuracy-prompt literature points the same way, modest, reliable improvements in what gets shared when accuracy is made salient at the moment of sharing (Pennycook et al. 2021; replicated at about half the effect by Roozenbeek et al. 2021). The mechanism is not asking people to do a lot of research. It's breaking their illusion of knowing: you discover you *don't* know, and that changes what you do.
How to do it.
-
Ask this one question, thumb over the button. If the honest answer is "I wouldn't," that's the answer.
-
Spend this pause on things you're seeing for the first time. Fazio found the effect roughly twice as large on novel headlines as on ones already seen once. By the third viewing it already sounds right.
Caution.
The effect is real and small (a third of a point on a six-point scale). It works at scale because sharing compounds; for any one message it's a nudge. And it only helps with content where your existing knowledge has something to say.
"Almost every successful manipulation, a phishing email, an outrage post, a dark pattern depends on the same thing: a reaction that happens before reflection can. My research asks whether we can train that missing pause."
Anna Collard, Founder, Digital Mindfulness
The Two Speeds:
Working with AI
What the evidence says.
AI shifted knowledge work from producing to checking, and checking is expensive: a BCG survey of 1,488 workers found supervising AI output associated with measurably more mental effort, fatigue and overload, and named the result "brain fry" (Bedard et al. 2026). Meanwhile scrutiny is being allocated by the wrong signal. The more people trust the tool, the less critically they think (Lee et al. 2025) and polish is exactly what these systems are best at.
How to do it.
Ask this question when the output lands: *what breaks if this is wrong?*
-
Nothing much → skim, without guilt.
-
My time → normal read.
-
My name → check mode: one thing open, read the load-bearing part properly, find at least one thing to question.
One hard rule above the feeling: anything external or irreversible gets check mode however trustworthy it looks. For the waits: under thirty seconds, just wait; longer, leave properly and do something you can *finish* the damaging pattern is hovering half-in two tasks (Leroy 2009).
Caution.
You will triage wrongly sometimes; the hard rule is the guard-rail. And most of what's frying people is structural: output expectations that rose the day the tools arrived. A sorting question won't fix that. The team version of this practice (see *For your team*) is better than the personal one.
"Almost every successful manipulation, a phishing email, an outrage post, a dark pattern depends on the same thing: a reaction that happens before reflection can. My research asks whether we can train that missing pause."
Anna Collard, Founder, Digital Mindfulness
The Five Minute Question
Interrupt Your Autopilot
What the evidence says.
Most phone-checking isn't a decision; a cue fires and the hand moves. That matters beyond distraction because automatic processing is where influence gets in. In a study of 472 people, an intuitive decision style was the strongest predictor of failing to tell real emails from fraudulent ones, and a rational style didn't independently help (Butavicius 2022). The authors concluded that interrupting the heuristic response may beat trying to summon analysis on demand.
How to do it.
Stop the hand mid-motion. Could this wait five minutes? If yes, phone face down. If no, pick it up deliberately, a legitimate outcome.You're building the noticing and a habit interrupt.
Caution.
Still busy with the research on that one.
"Almost every successful manipulation, a phishing email, an outrage post, a dark pattern depends on the same thing: a reaction that happens before reflection can. My research asks whether we can train that missing pause."
Anna Collard, Founder, Digital Mindfulness
Batching the Inbox
Three times per day instead of a constant feed
What the evidence says.
Kushlev & Dunn (2015) had 124 adults alternate between unrestricted email and three checks a day. The limited week produced significantly lower daily stress and less felt distraction. People found it hard to do, which is the point.
How to do it.
Agree the norm first (email is a four-hour medium), then set the windows. Notifications off at device level, not by willpower.
Caution.
Two weeks, self-report, mostly younger participants. And it only survives as a personal practice if your team's expectations allow it — which makes it a structural intervention in personal clothing.
"Almost every successful manipulation, a phishing email, an outrage post, a dark pattern depends on the same thing: a reaction that happens before reflection can. My research asks whether we can train that missing pause."
Anna Collard, Founder, Digital Mindfulness
Single Tasking
One thing at a time, for a protected interval.
What the evidence says.
Mostly indirect: divided attention and time pressure reliably degrade the ability to discriminate between similar things, which is the finding underneath most of what I study. The Pomodoro-style technique itself has little direct evidence; the conditions it protects do.
How to do it.
Make the block visible, treat it as a commitment, get the team to treat it as one. This works when the calendar protects it and fails within a fortnight when it's a private discipline against a culture of instant availability.
Caution.
It's a team agreement as much as it is a productivity method.
"Almost every successful manipulation, a phishing email, an outrage post, a dark pattern depends on the same thing: a reaction that happens before reflection can. My research asks whether we can train that missing pause."
Anna Collard, Founder, Digital Mindfulness
The Hourly Minute
Sixty seconds of movement, once an hour
What the evidence says.
The most solid evidence base here and the most modest claims. A meta-analysis of micro-breaks (Albulescu et al. 2022; 22 samples, 2,335 people) found reliable gains in vigour and reductions in fatigue but no significant effect on demanding cognitive performance, and longer breaks did more. An energy intervention, not a judgement intervention.
How to do it.
Two or three, sixty seconds each, cued by events (end of a call, sending something) rather than a timer.
-
Seated swimming at your desk: slow backstroke, opening the chest.
-
Forward fold, head heavy.
-
Spinal waves, arching and rounding through five breaths.
-
Shoulder rolls with eye rest, eyes closed or looking at the furthest thing in the room.
-
Stand and reach, one breath at the top.
Caution.
The vagus nerve is often invoked for why forward folds feel calming. The story is plausible; the evidence for specific postures is thinner than the confidence it's usually stated with. The movement helps. I'd leave the mechanism open.
"Almost every successful manipulation, a phishing email, an outrage post, a dark pattern depends on the same thing: a reaction that happens before reflection can. My research asks whether we can train that missing pause."
Anna Collard, Founder, Digital Mindfulness
Breath
Two or three rounds of box breathing before something that matters.
What the evidence says.
I'll be direct: the claim that a few minutes of breathing improves judgement has been tested and found null. A six-minute mindfulness induction before reading fake news changed nothing about how credible people found it, what it did is lower effort and arousal (Sebastião 2019). The most-cited workplace study on breathing breaks is a fourteen-person in-house EEG study. What breath reliably does is change arousal. That's worth having, but its only the first step, not the full protection or discernment.
How to do it.
Before joining a call, before opening something you're dreading, before replying. Then look at the message.
Take a deep breath into your nose through to your belly, count to 4 on the inhale.
-
Hold to the count of 4
-
Exhale to the count of 4
-
Hold to the count of 4.
Caution.
I use this daily. But the research so far is thin on it helping us protecting ourselves online. It sets the basis for a different state of being, but I need to do more research first.
"Almost every successful manipulation, a phishing email, an outrage post, a dark pattern depends on the same thing: a reaction that happens before reflection can. My research asks whether we can train that missing pause."
Anna Collard, Founder, Digital Mindfulness
The 1 to 3 Minute Opener
Meetings start with a short mindful opening
What the evidence says.
How to do it.
Every meeting starts with three shared minutes, on the agenda, owned by the chair.
Menu so it needs no preparation:
-
Three conscious breaths;
-
Two rounds of box breathing;
-
A five-senses grounding exercise: invite participants to focus their attention on 5 things they see, 4 things they can touch, 3 things they can hear, 2 things they can smell, and one taste.
-
Sixty seconds of silence: Invite everyone to close their eyes and notice their breath or feel their feet on the floor for 60 seconds.
-
The One-Word Check-in: Ask participants to share a single word describing how they arrive (such as tired, curious, or ready) without requiring a long explanation.
-
Intention Setting: Give the group 30 seconds of quiet time to define their personal goal or focus for the discussion
Caution.
Honest limit: the mechanism is that it's collective and scheduled, not that any exercise sharpens judgement. If leadership schedules over it, it doesn't exist. Message the practice, not the prevalence.
"Almost every successful manipulation, a phishing email, an outrage post, a dark pattern depends on the same thing: a reaction that happens before reflection can. My research asks whether we can train that missing pause."
Anna Collard, Founder, Digital Mindfulness
Meeting Time Defaults
Meeting times set to 25 or 45 mins by default.
What the evidence says.
The break evidence is modest (vigour and fatigue, not performance) but real.
How to do it.
Set the calendar system's default meeting lengths to 25 and 50 minutes, organisation-wide.
Zero individual behaviour change required, which is why it's the highest-leverage.
Caution.
Honest limit: the mechanism is that it's collective and scheduled, not that any exercise sharpens judgement. If leadership schedules over it, it doesn't exist. Message the practice, not the prevalence.
"Almost every successful manipulation, a phishing email, an outrage post, a dark pattern depends on the same thing: a reaction that happens before reflection can. My research asks whether we can train that missing pause."
Anna Collard, Founder, Digital Mindfulness
Response Time Norms
Email is a four-hour medium.
What the evidence says.
How to do it.
-
Email is a four-hour medium
-
Chat should not an interrupt channel for non-urgent items;
-
Nothing sent after hours expects a reply before morning (send-delay on by default).
This is what makes inbox batching possible for individuals. Without it, the personal practice fails in a fortnight.
Caution.
"Almost every successful manipulation, a phishing email, an outrage post, a dark pattern depends on the same thing: a reaction that happens before reflection can. My research asks whether we can train that missing pause."
Anna Collard, Founder, Digital Mindfulness
Decision Point Pauses on
Critical Actions
Friction design where it matters
What the evidence says.
How to do it.
Add specific warnings and procedural check at key points (payments, credential entry, external data sharing).
Not a generic "are you sure". Generic warnings make people more cautious without making them more accurate. A short, procedural check that names what to look at (sender domain, request channel, independent verification)
Procedural content is what builds discrimination rather than just raising suspicion (Sheng 2007; Xiong 2019 in the corpus).
Caution.
"Almost every successful manipulation, a phishing email, an outrage post, a dark pattern depends on the same thing: a reaction that happens before reflection can. My research asks whether we can train that missing pause."
Anna Collard, Founder, Digital Mindfulness
Quick Check and Feedback
Build how would I know if this were false or safe?
What the evidence says.
How to do it.
Make it easy in the workflow for people to ask, verify or double check if they have doubts.
Build how would I know if this were false? into the workflow:
a one-line check before external forwarding;
a standing "source?" reflex in team channels;
a named question in the comms approval chain.
This is the structural version of the best-evidenced individual practice in the catalogue (Fazio 2020; the accuracy-prompt literature). It targets false content without slowing true content.
Caution.
"Almost every successful manipulation, a phishing email, an outrage post, a dark pattern depends on the same thing: a reaction that happens before reflection can. My research asks whether we can train that missing pause."
Anna Collard, Founder, Digital Mindfulness
In the Moment Feedback
Without shame
What the evidence says.
Where simulations are used, feedback delivered at the moment of failure cut re-victimisation two weeks later from 50% to 40% in a preregistered trial (Bender 2026).
How to do it.
Caution.
Immediate feedback also reduced subsequent training completion and irritated people (Madeira & Alves 2026), and some simulation programmes are documented as inducing fear and shame. The tone is the design problem. Sanctions work briefly and then people reason they've "already been punished" (Kim 2020).
"Almost every successful manipulation, a phishing email, an outrage post, a dark pattern depends on the same thing: a reaction that happens before reflection can. My research asks whether we can train that missing pause."
Anna Collard, Founder, Digital Mindfulness
In the Moment Feedback
Without shame
What the evidence says.
Reviewing AI output is real cognitive work and is currently invisible in every workload model. Fatigue was lower where AI use was deliberately managed and integrated into shared workflows rather than adopted individually (Bedard et al. 2026).
How to do it.
-
Make AI review time a named, budgeted activity;
-
Agree what "under my name" means; label drafts as drafts.
-
Cap tools where you can.
Caution.
"Almost every successful manipulation, a phishing email, an outrage post, a dark pattern depends on the same thing: a reaction that happens before reflection can. My research asks whether we can train that missing pause."
Anna Collard, Founder, Digital Mindfulness
Closing thought..
This catalogue grows as the research does. When something here moves because a study lands, or because it turns out not to hold up. I'll say so on the page rather than quietly editing it.