Almost nobody decides, in a quiet moment, to be dishonest. The hard cases don't feel like a choice between right and wrong at all, they feel like a choice between right and a deadline, right and a number, right and the boss who is staring at you. That gap, between the values you hold when nothing is at stake and the values you act on when something is, is where integrity actually lives.

The quick version

  • Integrity is consistency between your values and your actions when the two come under pressure, a thing you do, not a personality badge you either own or don't.
  • Good people drift not by deciding to lie but through ethical fading: the moral stakes quietly drop out of the frame, so the call stops feeling like an ethical one at all.
  • What sets the standard is the leader's behaviour, not the values poster. People learn what's acceptable by watching what the powerful person does, especially under stress.
  • The defences that actually hold are practical rather than heroic: name the ethical stakes out loud, pre-script your response before you need it, and build small moments of friction that slow a fast, self-justifying decision down.

The idea in depth

Start with a definition that does some work. The most-cited empirical account of ethical leadership comes from Michael Brown, Linda Treviño and David Harrison, who defined it as "the demonstration of normatively appropriate conduct through personal actions and interpersonal relationships, and the promotion of such conduct to followers through two-way communication, reinforcement, and decision-making" (Brown, Treviño & Harrison, 2005, Organizational Behavior and Human Decision Processes). The phrase that matters is "personal actions." Their grounding is social learning theory: people work out what is acceptable here by watching a credible, powerful model, and the leader is the most-watched model in the room.

An earlier paper by the same lab, Treviño, Hartman & Brown (2000) in California Management Review, splits the job into two pillars. There's the moral person, your own honesty, fairness and concern for others, and the moral manager, the active work of making ethics visible through what you reward, what you talk about, and what you model. Their unhappy finding: being a good person privately isn't enough. A leader who is decent but silent on ethics tends to be read as ethically neutral, and a neutral signal in a high-pressure environment is taken as permission. So the move is to be loud about the standard before it's tested. Say, in the calm, what you expect when the numbers are ugly, because your team is calibrating off you, and silence calibrates downward.

One honest limitation here: this is correlational, reputational research. It tells us how leaders earn a reputation for integrity and how that reputation tracks with team outcomes; it can't promise that any single brave act produces a good result. Sometimes doing the right thing genuinely costs you, with no offsetting reward. Treat the evidence as a strong case for the habit, not a guarantee about any one instance.

Why good people drift

The more useful question isn't "why are some people corrupt", it's "why do honest people do things they'd be ashamed of, and not notice at the time." Two bodies of work answer it.

The first is Ann Tenbrunsel and David Messick's ethical fading (Tenbrunsel & Messick, 2004, Social Justice Research). Their argument: the moral dimension of a decision doesn't get overridden so much as it quietly drains out of the picture. Re-label "lying to the customer" as "managing expectations," frame a falsified figure as "an aggressive assumption," and the choice stops registering as an ethical one. You're not winning an argument with your conscience; you've stopped having the argument. So the move is to put the moral word back into the room. Ask, plainly, "is this honest?" rather than "is this within policy?" Naming the stake re-engages the part of judgement that fading switches off.

The second is Albert Bandura's moral disengagement, the mental manoeuvres that let people act against their own standards while still feeling like good people (Bandura, on moral disengagement). The ones that show up most at work are displacement of responsibility ("I was only following the target"), diffusion of responsibility ("everyone signs off, so no one does"), euphemistic labelling (the sanitising language above), and advantageous comparison ("it's nothing next to what the competitor pulled"). Each is a small permission slip. The move is to learn the four phrases as warning lights: when you hear yourself or your team reaching for one, that's the cue to slow down, not speed up.

People rarely choose to be dishonest. They choose a deadline, a number, a boss, and call it something other than what it is.

The limitation worth flagging: this research explains the slide better than it predicts the individual. Not everyone who uses a euphemism is mid-betrayal, and over-policing language turns a team paranoid. Use these as diagnostic prompts for your own judgement, not as a lie detector you point at colleagues.

flowchart TD
    A(["Pressure arrives: a deadline,
a number, a powerful ask"]) --> B(["Is the moral stake
still named out loud?"])
    B -->|"Yes, "is this honest?""| C(["Judgement stays engaged;
real choice is visible"])
    B -->|"No, re-labelled as
"policy" or "assumptions""| D(["Ethical fading:
the call stops feeling ethical"])
    D --> E(["Disengagement phrases:
"only following the target",
"everyone signs off""])
    E --> F(["Drift, and a clear
conscience the whole way"])
    C --> G(["Hold the line
(or escalate it)"])
How an honest person drifts without ever deciding to. Naming the moral stake is the fork. Leaders Loop

So what do you actually do?

The intuition that integrity is courage, that you just need to be braver, is mostly wrong, and unhelpful besides. Mary Gentile's Giving Voice to Values work reframes it usefully (Gentile, 2010). Her starting premise is that most people already know what's right and already want to act on it; the failure point is not knowing how to do it effectively in the moment, against real pushback. So she shifts the question from "what is the right thing?" to "how do I say and do the right thing well?", and her core method is pre-scripting: rehearsing your response to the dilemma you can see coming, before you're in it. This connects to two adjacent Toolkit ideas: holding the line is a form of consistency of word and action, and pre-scripting is impossible without the self-awareness and emotional self-regulation to catch yourself mid-rationalisation.

Three practical moves, each tied to the research above:

1. Name the stake (defeats fading). When a decision starts getting re-labelled, say the plain version out loud: "So we'd be telling the client something we know isn't true." It's almost rude, which is the point, it drags the moral dimension back into a frame that was quietly dropping it.

2. Pre-script the hard line (Gentile's move). Decide your words for the predictable pressure before it lands. "I can't put my name to that number, but here's what I can stand behind" is easier to say when it's the second time you've said it, even if the first time was only to yourself in the car.

3. Build friction (slows disengagement). Disengagement thrives on speed. A 24-hour rule on a decision you're uneasy about, a named devil's advocate, a "would I be comfortable explaining this to my team?" gate, each is a small piece of grit that gives engaged judgement time to catch up with the fast, self-justifying one.

flowchart LR
    P(["Predictable
pressure point"]) --> N(["NAME
the moral stake plainly"])
    N --> S(["SCRIPT
your line in advance"])
    S --> F(["Add FRICTION
time, a challenger, a test"])
    F --> H(["Act in line
with your values"])
    H -.->|"reflect after,
refine the script"| P
Name, script, friction: the loop that turns "be braver" into something you can actually run. Leaders Loop

A worked example

The figures below are illustrative, a composite, not a real company.

Priya runs delivery for a mid-size software team. It's the last week of the quarter and a feature the sales team has already promised three customers isn't ready: a payments edge case still fails about one in forty times. Her VP wants it shipped Friday. "Mark it done," he says. "We'll hotfix the edge case next sprint. Nobody's going to hit it."

Watch the fading start. "Mark it done" is cleaner than "report a known defect as resolved." "Nobody's going to hit it" is advantageous comparison, the 1-in-40 becomes a rounding error. "We'll fix it next sprint" displaces responsibility onto a future that may never arrive. Each phrase is reasonable on its own. Together they're a slide, and the VP almost certainly doesn't experience himself as asking Priya to lie.

The brave-but-vague version of Priya pushes back on willpower, gets told she's not a team player, and either folds or burns a relationship. The version who has done the Toolkit work runs the three moves instead. She names the stake: "Just so we're saying it plainly, we'd be telling three customers a known payments bug is fixed when it isn't." She uses a pre-scripted line: "I can't mark it resolved. What I can do is ship it flagged as a known limitation with a fix date, and I'll own that message to the customers." And she's built friction into how her team works, so there's a standing rule that "resolved" means tested, which makes her line the policy talking, not just her.

Notice what changed and what didn't. She didn't out-argue her VP on ethics or deliver a speech about honesty. She made the real choice visible, offered a route that protected both the customer and the deadline, and leaned on a rule that existed before the heat. That's integrity as a system, not a stand. Sometimes it still costs her. The research can't promise the VP says yes, only that this is how the standard holds, and how a team learns what "done" really means.

Frequently asked questions

Isn't integrity just a personality trait you either have or you don't?

That's the comfortable version, and it lets everyone off the hook, the honest are praised, the rest are written off. The evidence points elsewhere: ordinary, well-intentioned people act against their values under specific, predictable conditions (Tenbrunsel & Messick's fading; Bandura's disengagement). If situations do most of the bending, then integrity is partly an engineering problem, design the conditions, don't just audit the character.

How is this different from just having good values?

Values are what you'd choose with nothing at stake. Integrity is the match between those values and what you actually do when something is at stake. Gentile's whole point is that the gap is rarely about not knowing right from wrong; it's about not having a workable way to act on what you already know, under pressure. Knowing isn't the bottleneck, doing is. (For the values layer underneath, see values-based leadership.)

What if doing the right thing genuinely damages my career?

Sometimes it will, and pretending otherwise is dishonest. Two honest things, though. First, most workplace integrity tests are smaller and more survivable than the dramatic whistle-blower story, a flagged figure, a "this isn't ready" said early. Pre-scripting and friction handle these without heroics. Second, for the genuinely costly cases, that's a decision only you can own, but you'll own it better having seen the choice clearly than having let it fade. This isn't legal advice; if a situation involves potential illegality or regulated duties, get qualified counsel for your jurisdiction.

How do I build this on my team without sounding preachy?

Skip the values poster, silence on ethics reads as neutrality, but lecturing reads as theatre. Treviño and colleagues' "moral manager" finding is about visible behaviour: what you reward, what you tolerate, what you do yourself when it's expensive. Define your standards in concrete operational terms ("'done' means tested"), and the message lands without a sermon. It's role-modelling more than messaging, see role-modelling and leading by example.

I caught myself mid-rationalisation. Now what?

That's the skill working, not failing. Noticing a euphemism leave your own mouth, "managing expectations," "aggressive assumption", is exactly the warning light. The move is to stop and re-state the plain version, then check whether you'd still make the call. Catching the slide is most of the battle; almost nobody who's genuinely paused to ask "is this honest?" proceeds to lie anyway.

Related in the Toolkit

Where to go next