Sometime in the middle of my freshman year, at the advice of a friend, I took a walk to the local dollar store and bought myself a calendar style daily planner. For years before then, I had relied entirely on a to-do list on my phone, adding and removing obligations as needed. It turned out to be a major improvement; I missed significantly fewer deadlines and it gave me the opportunity to get a general idea of what to expect from each coming week. Naturally, at the beginning of my sophomore year, I decided to take the next logical step and start scheduling hour-by-hour, at which point I promptly lost all the progress I had made. The problem was that, even when I added ‘buffer time’ to compensate, I still found myself underestimating the time I needed for anything that didn’t have a built in beginning and end, like classes and meetings.
Even this blog post was underestimated. A few days ago, I told myself I could definitely start and finish on Saturday night, leaving myself all of Sunday night to study for my exam at noon on Monday. How did things go so wrong? We might begin to answer that question by considering this explanation from a paper on intuitive prediction by Daniel Kahneman and Amos Tversky, which distinguishes between singular and distributional data. The former, according to the authors, is based on the various factors which characterize a scenario, while the latter is based on comparisons with other similar scenarios. In other words, as applied to homework, a singular approach might be based on the time needed to choose a topic, conduct my research, pull useful information from my sources, and so on. A distributional approach, by contrast, would draw on my prior experience in similar assignments (including the last two blog posts), as well as any information I happen to have about how long it takes my peers.
Kahneman and Tversky identify the over-reliance on internal models (singular data) as the primary component of what they called ‘the planning fallacy’. Even disregarding the complementary effects of wishful thinking, it seems that even experts tend to create unreasonable time estimates based on a preference for internal rationalizations. The article reasons that external (distributional) assessments tend to yield more accurate results because they account for the multitude of possible complications which would otherwise be unaccounted for on the basis of individual improbability.
A research paper from 1994 by Buehler, Griffin, and Ross built on this dichotomy. Their first two studies concerned the completion of an Honors Thesis project and either an academic or personal task, respectively. Earlier predictions were correlated with earlier completion times but, in all cases, participants took consistently longer than they had planned despite reporting high levels of confidence in their numbers and even when asked to give a pessimistic estimate. In fact, the best predictor available turned out to be the externally established deadline for a project’s completion, although time-to-deadline was shown to have minimal influence on estimation.
The third study in the same paper asked participants to reason aloud while making their predictions, showing that future plans are disproportionately favored over all other considerations, although this study failed to support the researchers’ hypothesis that those who favored past experiences would generate more accurate estimates. However, the subsequent study was able to improve guesses under the specific case in which subjects were not only asked to consider similar past experiences, but were also forced to describe how said experiences were relevant to the future they were planning for.
Perhaps most interesting was the fifth and final study, in which observers were brought in and asked to make their own independent estimates for other participants after being given necessary information about the person they were assessing and the parameters of the project. These social predictions were consistently more conservative, with a day and a half on average more time given for one-week deadlines and over four days greater for two-week deadlines when compared to the individual predictions of those actually performing the task. Observers, as predicted, were more likely to use impersonal distributional data and were thus more accurate, just as with the prior task in which subjects were manipulated to relate their plans to past actions.
So what can we do differently to improve planning in our own lives? First, it is important to consider the task at hand in relation to similar tasks from past experience, which requires overcoming the mental bias towards viewing past complications as relatively externally caused and unique. Try to find ways in which the upcoming situation is specifically similar to what you already know about yourself and others, which will help to decrease the effects of over-optimism. Perhaps an even more practical approach is to outsource the planning entirely and enlist your friends to help. Due to their external perspective, they are far more likely to make estimates which prioritize their general knowledge over your best intentions, which leads to more realistic expectations.