HUMOR AT WORK
We built Reliefr for the gaps between your meetings.
You've seen the general science of humor. This is the workplace cut — the research on meeting fatigue, micro-recovery, and why a well-timed 30-second interruption is a legitimate intervention and not just a gimmick.
At a glance
Four stats. The rest of the page is how we got from these to the app you're reading about.
Part 1 · The problem
In 2021, Microsoft's Human Factors Lab ran a small EEG pilot on 14 knowledge workers doing back-to-back video meetings. The finding they reported: beta-wave activity — the brain signal associated with stress — climbed continuously as the meetings stacked. A 10-minute break between meetings returned that activity to baseline; no break, and stress simply accumulated through the day.1
A caveat the internet consistently ignores: this was a Microsoft internal pilot with N=14. It was never peer-reviewed. We cite it because the signal it describes lines up with a much larger body of peer-reviewed work on cognitive depletion and recovery — not because a vendor whitepaper is a gold-standard study.
The bigger number is stronger: Microsoft's 2022 Work Trend Index, drawn from telemetry across millions of Microsoft 365 users, showed that knowledge workers were attending roughly three times as many meetings per week as they had in February 2020.2 Gallup's 2024 global report adds the felt consequence — around 41% of workers worldwide say they experience stress "a lot of the day."3
You don't need an EEG to describe this. You already know what it feels like.
Part 2 · The 30-second case
The best peer-reviewed evidence for the format Reliefr uses — brief, interrupting, in-context — comes from micro-break research. A 2022 meta-analysis in PLOS ONE pooled 22 studies on breaks of ten minutes or less. Across roughly 2,300 participants, micro-breaks improved vigor and reduced fatigue with small-to-medium effect sizes (g ≈ 0.36 for vigor; similarly-sized reductions in fatigue). The effects were strongest for cognitive and clerical tasks — knowledge work.4
A 2018 randomized controlled trial of 153 knowledge workers found that short lunchtime walks in a park or brief relaxation exercises reduced end-of-workday fatigue and strain. The effects held across a two-week study period — this wasn't a one-day lab artifact.5
A notification isn't a park walk. But a 30-second laugh between two tense meetings is closer to a micro-break than it is to nothing — and in this literature, the gap between "nothing" and "a small well-timed thing" is the one that matters.
Part 3 · The humor-specific evidence
A 2012 meta-analysis in the Journal of Managerial Psychology, synthesizing roughly 49 samples on workplace humor, reported a consistent pattern: positive humor at work was negatively related to burnout and stress, and positively related to work performance. The individual effect sizes were modest, but the direction was stable across studies.6
A 2017 paper in the Journal of Organizational Behavior looked specifically at affiliative humor from supervisors. Where it was present, subordinates reported lower emotional exhaustion — a core dimension of burnout. The effect was contingent on context (some employees respond better to humor than others), which is itself useful information: humor doesn't replace judgment about when and how.7
Neither of these is an RCT proving humor causes anything in a specific workplace. Causal evidence in workplace humor research is sparse — most studies are correlational or lab-based, and the field is honest about that. What we have is a consistent directional signal across dozens of independent studies, which is how most applied social science actually works.
Part 4 · Your best ideas are downstream of your mood
In a 1987 experiment now considered a classic of the field, Isen and colleagues showed participants either a short comedy clip, a neutral film, or nothing, then gave them Duncker's candle problem — a well-known test of creative insight. Subjects primed with the comedy clip solved the problem at much higher rates than the control groups.8 The study has been replicated and extended many times.
The workplace version of that finding comes from Amabile and colleagues at Harvard, who ran a longitudinal diary study across 222 employees in seven companies. Their finding: positive affect reliably preceded creative thought — often by a day or two. The ordering mattered. Mood wasn't downstream of creativity; creativity was downstream of mood.9
The broader theoretical frame comes from Barbara Fredrickson's broaden-and-build theory — the idea that positive emotions expand attention and build durable psychological resources over time, rather than just feeling nice in the moment.10
Translation: if your morning starts in a hole, the ideas you'll have that afternoon are already compromised. Small inputs compound in the direction of whichever mood you're running in.
Part 5 · Teams that laugh together
In 2015, Google published the results of Project Aristotle, a two-year internal study of about 180 Google teams looking for what separated high-performing teams from everyone else. The answer, surprisingly clean for an organizational-behavior finding, was psychological safety — the shared belief that it's safe to take interpersonal risks on the team.11
The academic version of this work was already old by then. Amy Edmondson published the foundational paper on psychological safety in Administrative Science Quarterly in 1999, and a 2017 meta-analysis in Personnel Psychology confirmed that psychological safety consistently predicts team learning, performance, and engagement across hundreds of studies.1213
Humor is one of the observable surface signals of psychological safety. A 2014 Journal of Applied Psychology study by Lehmann-Willenbrock and Allen videotaped 54 team meetings and coded their humor patterns. The humor-support patterns they identified predicted supervisor-rated team performance two years later.14 Not a short-run effect. A real one.
One honest caveat: humor is a double-edged signal. Bitterly, Brooks, and Schweitzer's 2017 work, across eight studies with about 2,200 participants, showed that successful humor raises perceived status and competence — and failed or inappropriate humor lowers them. Quality matters.15
Part 6 · The ghost at the standup
Absenteeism is the metric every HR dashboard tracks. Presenteeism — the phenomenon of showing up physically but being cognitively checked out, depleted, or unwell — is the one that eats the actual budget. Paul Hemp's 2004 Harvard Business Review piece laid out the case plainly: for many U.S. employers, the cost of presenteeism exceeds the cost of absence. Goetzel and colleagues put numbers to it in a widely-cited JOEM review the same year, estimating substantial per-employee losses from conditions including depression, stress, and fatigue that keep people at their desks without functioning.1617
The engagement side of the same coin comes from Gallup's Q12 meta-analysis — their 2020 technical report, pooling 456 studies across 2.7 million employees in 276 organizations. Top-quartile engaged business units showed 81% lower absenteeism and 18–23% higher productivity than bottom-quartile units. Not a claim about a specific intervention; a claim about the direction affect moves the org.18
Reliefr is not claiming to fix presenteeism. What we're claiming is that the gap between "depleted but present" and "actually functional" is narrower than most wellness marketing will admit — and it's the gap small, well-timed interventions can live in.
Our principles at work
Reliefr's humor darkness slider goes from wholesome to unhinged. The research is clear on something our content rules already reflect: affiliative humor (humor that builds connection) is associated with positive wellbeing outcomes. Humor that disparages specific people is associated with worse ones. We never do the latter — but we absolutely do edge.
So we calibrate by context, not just by darkness level — but we don't hardcode what "the right mood" is for each context. The algorithm learns from your ratings which moods actually work for you in each moment. One user might want commiserating ("ugh, same") during stress; another wants cathartic ("yes, finally someone said it"). One pre-meeting moment calls for gallows-edge; another wants something soft to take the pressure off. The app figures out which is which, per user, per context — and shows its work: long-press any joke for a "Why this joke?" sheet that lays out the reasoning. Always aimed at the situation, never at a person.
You control the darkness. The algorithm learns the direction.
The anti-pattern
There's no feed to scroll, no "just one more" joke, no engagement loop built to keep you in the app. A joke arrives at a moment it's likely to land — 15 minutes before a hard meeting, when your heart rate runs hot, the morning after a rough night — and then the app steps aside. That 30-second interruption exists to give you back to the day, not take more of it. Work is part of that day. So is lunch, your kids, the walk home, and whatever else you're actually living for.
The research on what an always-on attention-seeking app does is unambiguous. Ward et al. (2017) showed that the mere presence of one's own smartphone on the desk measurably reduced available cognitive capacity on a working-memory task — without the phone ever being used.19 Satici et al. (2023) found that doomscrolling — the passive, endless consumption of distressing content — is associated with higher psychological distress and lower subjective wellbeing, independent of total social media use.20 Meanwhile the micro-break literature keeps finding the opposite: brief, bounded recovery moments help.4
We built around the second finding and explicitly reject the design patterns of the first.
Timed relief, not infinite content.
Integration
We didn't reverse-engineer the research to justify the app. The app was built around these studies from the start.
When your wearable's heart rate runs hot against your baseline, Reliefr fires a joke chosen for the moment. The algorithm learns from your ratings which mood actually calms you down — for some users that's commiserating, for others cathartic or wholesome. Maps to the humor-and-burnout meta-analytic evidence67.
After a poor night of sleep (detected via Health Connect), Reliefr adjusts the morning delivery — picking whatever the algorithm has learned helps you start the day on the days you needed extra sleep. Maps to Fredrickson's broaden-and-build work10 and the Amabile finding that positive affect precedes creative and productive output by up to two days9.
Jokes are filtered by industry / age / gender / humor darkness, then scored on three independent dimensions the algorithm learns per user: topic (what the joke is about), style (how it's constructed), and mood (the emotional residue) — each tracked separately for each delivery context. Humor landing depends on shared reference frames, which is exactly what Bitterly et al. warn about when humor fails15. Long-press any delivered joke for a "Why this joke?" sheet that shows the picker's reasoning.
Caveats
Workplace humor RCTs are rare. Most of the evidence on this page is correlational, lab-based, or from adjacent literatures (positive affect, micro-breaks, psychological safety) that we're bridging into humor. The directional signal is strong; individual effect sizes are modest.
The Microsoft EEG pilot is a Microsoft whitepaper, not peer-reviewed. We cite it because it's the evocative one — but we pair it with the Albulescu 2022 meta-analysis, which is the one that actually holds up.
Humor can fail. Bitterly et al. are clear: humor that misses lowers status and trust. Our answer is curation, not volume — calibrated by context, reviewed for direction, never a joke aimed at a person.
We are not a therapy replacement or a clinical tool. If you're struggling with burnout, depression, or workplace harassment, talk to someone qualified. A notification is not a care plan.
The research on laughter, cortisol, dark humor, and humor styles lives on its own page.
Sources
Every superscript on this page links here. Where possible, peer-reviewed primary sources are preferred over journalism and vendor whitepapers; where we rely on the latter, we label it.