Plans that survive the moment

Lesson 3 left you with a problem and two repairs that do not work. Padding the estimate moves the number without improving it, and it is the first thing anybody tries. Being told to remember your history moves almost nothing at all.

This lesson is the third condition of that same study, which works in one specific way, and then the one planning technique in this whole subject with a meta-analysis behind it. Both come with limits that matter more than the headline, so they're in the lesson rather than in a footnote.

The third condition

The computer-assignment study had 123 Canadian university students and three groups, published in 1994, and lesson 3 showed you two of them.¹

The third group did what the recall group did, describing their past experience with similar assignments. Then they answered two further questions, and those two questions are what separates this group from the last one. The paper describes them rather than printing them; put into the second person they are:

• When would you finish this assignment if you completed it as far before its deadline as you typically do?
• What is a plausible scenario, based on your past experience, that would result in you finishing at your typical time?

Read the middle row before the bottom one.

The work didn't speed up. It took 7.0 days against the control group's 6.8, which if anything is a shade slower. What is different is the prediction: this group's sat at 7.0 against the control group's 5.5, which is where what actually happened was all along.

The share finishing on time roughly doubled against the control group, and it doubled because the forecast came to the work rather than the work coming to the forecast.

What it bought, and what it did not

The authors are careful about this and so must the lesson be: "Predictions were no more accurate in the recall-relevant condition than in the other conditions."¹

Look at what that's saying, because it's the same sentence lesson 3 met about padding. The predictions were no closer to the truth on any individual job. The spread stayed. What changed is that they stopped being systematically early.

There is one more figure from that study worth carrying, because it says how hard the connecting is. Even inside the group made to do it, only 12.5 percent of subjects mentioned their own past experiences when asked what they had been thinking about.¹ Adding those two questions to the recall the middle group had already done took the on-time share from 38.1 percent to 60, and most of the people answering them still did not report thinking about their history.

Remembering is not connecting, which lesson 3 stopped on, and connecting is not natural either. In the study it had to be asked for, in writing, and this course's reading is that the same is true outside it.⁴

Worked: the drill on a real job

Take a job you keep meaning to do. The worked case here is a constructed written report,⁴ and the drill runs the same on a tyre change or a tax return.

Step one, the ordinary estimate. "About three evenings." Write it down first, because you want to see what it was.

Step two, the history, stated as a fact rather than as a feeling. Not "I usually run over", which is the sentence the recall group could produce without it changing a thing. The actual pattern: the last two reports were each finished the night before they were due, and the one before that was two days late.

Step three, the first question. If this goes as it typically does, when is it finished? Not when you'd like it finished. The report is due in a fortnight. The typical pattern is the night before. So: thirteen days.

Step four, the second question, which is the one people skip. What plausible run of events produces that? And this is where you have to write something specific rather than shrug.

Step five: the number you carry is the one from step three, not the one from step one. Thirteen days, not three evenings.

From knowing to doing, which is a different problem

A better estimate tells you when something will be done if you do it. It says nothing about whether you'll start.

Those are genuinely separate, and the size of the second one has been measured. Gollwitzer and Sheeran, writing in the chapter this course has read in full, report that goal intentions "accounted for 28% of the variance in behavior, on average, across 422 studies"; that a review of health behaviour matrices found people "translated their 'good' intentions into action only 53% of the time"; and that in experiments which successfully changed people's intentions, the resulting change in behaviour was "only small-to-medium (R² = .03)", R² being the same variance-explained figure How to Learn Anything lesson 7 took apart. Their conclusion is printed in bold in the original: "forming even strong goal intentions does not guarantee goal attainment."²

Twenty-eight percent is a real relationship. What you intend has a lot to do with what you do. It's also a long way from everything, and the gap is the subject of the rest of this lesson.

Two kinds of intention

The distinction is the whole technique, and both halves have a definition in the authors' own words.

A goal intention is "the instructions that people give themselves to perform particular behaviors or to achieve certain desired outcomes", measured "by items of the form, 'I intend to achieve X!'"²

"I'm going to get fitter." "I'll sort the paperwork out." "I want to read more."

An implementation intention is different in one specific way. They "are if-then plans that link situational cues (i.e., good opportunities to act, critical moments) with responses that are effective in attaining goals or desired outcomes ('If situation Y is encountered, then I will initiate behavior Z in order to reach goal X!')."²

"When I have finished loading the dishwasher, then I will do fifteen minutes of the paperwork."

The difference isn't that the second is more detailed or more serious. It is that the second one names a situation. The first waits for you to decide; the second is attached to something that will happen anyway.

That is what the form is doing, and in this course's reading it is why the "when" half matters more than the "then" half,⁴ which is the reverse of how most people write them.

Worked: one goal, written twice

Dev is constructed.⁴ He wants to keep up with the professional reading his trade body puts out, and has wanted to for about a year.

As a goal intention, which is how he has been carrying it: I intend to keep up with the reading.

There's nothing wrong with that sentence. It's honest, he means it, and it's produced nothing in a year, which is the 53 percent and the 28 percent above arriving in one person's life.

As an if-then plan, first attempt: When I have a free evening, then I will read one article.

Now apply the cue test. Would he notice "a free evening" while tired and in a bad mood? No, because a free evening isn't an event. It is a judgement he has to make about an evening, and making that judgement is exactly the thing that fails on the evening it matters.

Second attempt: When I sit down on the train on Wednesday morning, then I will read one article.

That is an event. It happens whether or not he is thinking about the plan, he can't miss it, and it arrives at a time of day he's capable of reading. Notice what changed between the two versions: nothing about the goal, nothing about his willingness, and nothing about the "then" half. The whole repair was in the "when".

One more thing worth saying, because it is the honest limit on the example. Dev's plan might still fail. If it does four Wednesdays running, on a cue that good, the next section is what to check.

What the meta-analysis reports

The pooled result of their 2006 meta-analysis is d = .65, across 94 independent tests and more than 8,000 participants.³

Two things about that number before anything is done with it.

What d means. It is the gap between two groups measured in standard deviations of the outcome rather than in any real unit, which is what lets results on different outcomes be pooled at all. By the convention usually applied to it, .65 sits between the values called medium and large. So: a real effect, comfortably bigger than nothing, and nowhere near what the genre promises.

What this course has read. The chapter in which the authors state the idea and the numbers above was read in full. The meta-analysis itself was not opened, and the figures in this paragraph are at the level of a search summary. This course names the gaps in its own evidence, as lesson 1 did with the shape of its samples; this is a gap in what has been read: treat d = .65 as what the authors report rather than as something checked here.

The condition attached to the effect

This is the sentence that has to travel with the number, and it is from the same page as the definition:

Implementation intention effects "are stronger when self-regulatory problems beset goal striving, and when if-then planning is supported by strong activated goal intentions."²

Read the second half slowly. An if-then plan is a bridge from wanting to doing. It does nothing whatever about whether you actually want the thing.

So the technique has a shape. It helps most where you genuinely hold a goal and keep failing to act on it, which is a large and frustrating category. It helps least where you're carrying a goal you don't actually want, which is a different and larger category still, and lesson 6 is about that one.

Worked: the plan that was never going to work

Marek is constructed, and so are his four weeks.⁴ He wrote a good plan. "When I sit down after supper on Tuesday, then I will do half an hour of Spanish." Cue named, an event rather than a mood, a specific response attached to it.

Four weeks later he has done it twice.

Run the checks in order. Is it in the form? Yes. Is the cue an event he would notice while tired? Sitting down after supper on a Tuesday is about as concrete as a cue gets. Is the response small enough? Half an hour is not heroic.

The form isn't the problem, and this is where most advice stops and tells him to try harder.

The question the research points at is the goal behind it. Marek doesn't want to speak Spanish. He wants to have spoken Spanish, which is a different thing, and he took up the goal because a friend did. The condition in the quotation above is doing real work here: if-then planning is supported by strong activated goal intentions, and there is no strong goal intention here for it to be supported by.

The honest outcome of Marek's four weeks isn't a better plan. It is crossing the goal off. That is a legitimate result and this course is going to say so plainly rather than treating every abandoned intention as a failure of method.

And the diagnostic generalises, with one branch that has to travel with it. A plan that fails in the right form, on a good cue, four weeks running, is evidence about the goal rather than about the plan.

Which doesn't always mean the goal should go. Some goals aren't yours to drop: the paperwork, the errand somebody depends on, the thing your job requires. For those, "do I hold this goal" has an answer you already know without asking, and the authors' other half is the relevant one, that the effect is stronger where a self-regulatory problem is what stands between you and something you are going to have to do anyway. Marek's answer is to cross it off. A reader whose goal is imposed has a different question, which is what lesson 6 is about.

Four things people get wrong about this

"An if-then plan is just a goal with extra words." The extra words are a situation, which is the part that does the work. "I'll do the paperwork" waits for you to decide, forever if you let it. "When the dishwasher is loaded, then fifteen minutes of paperwork" is attached to a thing that happens whether or not you decide anything.

"d = .65 means it will work for me." An effect size is an average over 94 tests and more than 8,000 people.³ It says something about the technique and nothing at all about your Tuesday. The same warning Using AI Effectively lesson 11 gave you about capability claims applies exactly here: a pooled figure describes a body of studies, and the only measurement of you is one you take.

"This is willpower." The technique is aimed precisely at the case where deciding again in the moment is what keeps failing. Whether it removes the deciding or merely moves it earlier is a question about mechanism that this course has not read the evidence on, and it does not change what to do.

"So planning works." That overshoots. What is reported is a medium-to-large pooled effect on goal attainment for if-then plans,³ under a condition the authors state elsewhere.² "Planning works" is a claim about a whole genre, and lesson 8 is about what that genre can and cannot show.

Practice

Connections

Back. Lesson 3 set up the problem and ruled out padding and remembering; this lesson is the condition of that same study which worked, and it inherits the study's scope. Lesson 3's distinction between bias and accuracy is the one doing the work in the "what it bought" section, and it is worth noticing that the same distinction settles two different results in two lessons. Using AI Effectively lesson 11 gave you the habit of refusing to read a pooled figure as a prediction about yourself.

Forward. Lesson 5 is why the day you plan into does not hold still, which is the main reason a good plan meets a bad afternoon. Lesson 6 is about the goals that should be crossed off rather than planned around, which is where Marek's case ends up. Lesson 7 puts the plans and the ratios into an actual week. Lesson 8 asks what the whole evidence base, this technique included, is worth.

Go deeper

• Exploring the "Planning Fallacy" (Journal of Personality and Social Psychology, 1994). This course has read the abstract, Studies 1 and 2 in full, and Study 4 with its table and discussion, which is the study this lesson opens on. Study 4's table is the one to look at, since it holds all three conditions side by side.
• Implementation Intentions, Gollwitzer and Sheeran's own short statement of the idea, posted by the US National Cancer Institute. Four pages, free, and this course has read them. It is the source of every quotation in this lesson about what the two kinds of intention are.

Sources

1. Roger Buehler, Dale Griffin and Michael Ross, "Exploring the 'Planning Fallacy': Why People Underestimate Their Task Completion Times", Journal of Personality and Social Psychology 67(3), 1994, pages 374 to 376. Read in substantial part; Study 4 read with its table and discussion. Supports: the three conditions and the two questions put to the recall-relevant group; the predicted, actual and on-time figures for all three columns; the quoted sentence that predictions were no more accurate in the recall-relevant condition; and the 12.5 percent who mentioned past experiences within that condition. Scope: 123 university students in Canada on one computer assignment, published 1994.
2. Peter M. Gollwitzer and Paschal Sheeran, "Implementation Intentions", posted by the US National Cancer Institute, pages 1 to 4. Read in full. Supports: the definition of an implementation intention and of a goal intention, both quoted; the 28 percent of variance across 422 studies, the 53 percent translation rate, and the R² of .03; the bolded conclusion that forming even strong goal intentions does not guarantee goal attainment; and the quoted condition that effects are stronger when self-regulatory problems beset goal striving and when if-then planning is supported by strong activated goal intentions.
3. Peter M. Gollwitzer and Paschal Sheeran, "Implementation Intentions and Goal Achievement: A Meta-Analysis of Effects and Processes", Advances in Experimental Social Psychology 38, 2006, pages 69 to 119. Search-summary level only; this course has not opened the meta-analysis, and the body says so where the figure appears. Supports, at that level: d = .65 across 94 independent tests and more than 8,000 participants.
4. The reading of d as a gap in standard deviations, and the convention that .65 lies between medium and large, is stated as a convention rather than as a finding, and no source in research/SOURCES.md is cited for it. Marek is constructed, and so is the report in the worked drill; no source read for this course describes an individual's plan.