Bias in memory predicts bias in estimation of future task duration

Abstract

It often takes people longer than planned to finish tasks. Even though people are aware that their estimates have come up short in the past, they continue to underestimate the duration of future projects. Studies have shown underestimation for such tasks as writing papers, performing everyday and school tasks, completing a computer assignment (Buehler, Griffin, & Ross, 1994;Connolly & Dean, 1997;Griffin & Buehler, 1999;Koole & Spijker, 2000;Newby-Clark, Ross, Buehler, Koehler, & Griffin, 2000;Taylor, Pham, Rivkin, & Armor, 1998), building a computer stand (Byram, 1997), doing a spell check task (Francis-Smythe & Robertson, 1999), completing tax forms (Buehler, Griffin, & MacDonald, 1997), finishing Christmas shopping (Buehler & Griffin, 2003), completing Web development projects (Moløkken & Jørgensen, 2005), programming software (Jørgensen & Sjøberg, 2001), reading a manuscript (Josephs & Hahn, 1995), and waiting in line for gas (Konecni & Ebbesen, 1976).In the business world, underestimation of task duration has numerous negative consequences: missed deadlines, budget overruns, terminated contracts, and loss of business. These negative effects of task underestimation may be quite large in the multimillion dollar software design business (Connolly & Dean, 1997).The memory bias account (Roy, Christenfeld, & McKenzie, 2005) supplies a possible explanation for this tendency to underestimate future duration: Our memories for past durations are incorrect, causing corresponding errors in prediction. People remember past task durations as being shorter than they actually were, and therefore, when they consult them to predict the next task, they underestimate how long that task will take.Memories of past task durations will generally be based on estimates, rather than on actual measured durations, since it is rare that people know the exact beginning and ending times of tasks, especially when the task is completed over a number of different sessions. Studies of software development companies have shown that the people in charge of predicting how long projects will take rarely know the actual duration for previous, similar projects (Jørgensen & Sjøberg, 2001;Moløkken & Jørgensen, 2005). Furthermore, research has indicated that there is a tendency to underestimate task duration retrospectively (for reviews, see Block & Zakay, 1997;Fraisse, 1963;Poynter, 1989;Wallace & Rabin, 1960). If people rely on estimations and estimations tend to be short, predictions are likely to underestimate actual durations.It is unlikely that when people predict future duration, they recall every past instance of similar tasks and estimate the duration for each. At times, people may recall the duration of a specific instance, but they may also call upon a more general representation that they have for that task that is based on past experience. This past experience could be one of performing the task directly or observing others. A prediction then could be made by adjusting this general representation of duration for a task up or down on the ba...

Effect of task length on remembered and predicted duration

“…Short tasks are remembered as taking longer than they actually did, which in turn will cause overestimations of how long a short task will require in the future. Consistent with this idea, researchers have found that there is a similar effect of experience with a task on both remembered and predicted duration (Roy & Christenfeld, 2007) and that previous experience affects prediction for how long a task will take (Thomas, Handley, & Newstead, 2007). Further, it is not even clear how people could make a prediction of duration that is even moderately accurate without some reference to memories of how long similar tasks have taken in the past.…”

“…An index was created by taking the log of the ratio of estimated duration to actual duration in order to assess accuracy of estimates (Roy & Christenfeld, 2007). This index, which we refer to as log proportional error, has a number of benefits.…”

“…That is, when predicting how long a task will take, people may consult their memories of past completion times; thus, any bias in memory will simply transfer to prediction (Roy, Christenfeld, & McKenzie, 2005). For example, task novelty has a similar effect on bias for both remembered and predicted duration, with novel tasks likely overestimated and familiar tasks likely underestimated (Roy & Christenfeld, 2007). In contrast, previous explanations for bias in predicted duration-such as the planning fallacy (Kahneman & Tversky, 1982)-claim that bias likely occurs because of people's disregarding prior experience and focusing narrowly on the specific demands of the task at hand when estimating completion time.…”

Unpacking estimates of task duration: The role of typicality and temporality

“…Another account of the planning fallacy holds that underestimation occurs not because information about previous tasks is neglected but because it is inaccurately remembered (Roy, Christenfeld, & McKenzie, 2005). Whilst there is support for each account (e.g., Buehler et al, 1994;Roy & Christenfeld, 2007;Thomas & Handley, 2008), the research emphasis has been on calibration (Do predictions accord with actual and recollected duration?) and content (Do people focus on the unique features of the target task?…”

Belief in Optimism Might Be More Problematic than Actual Optimism

“…However, bias in remembered duration often appears to be due to task characteristics such as novelty (Boltz et al, 1998;Hinds, 1999;Roy and Christenfeld, 2007;Tobin et al, 2010), relative duration (Yarmey, 2000;Tobin and Grondin, 2009), size of potential estimation anchors (Thomas and Handley, 2008), and duration since completion . These tasks characteristics may alter attention that is paid to the task (Thomas and Weaver, 1975) or memory storage size associated with the task (Ornstein, 1969;Block and Reed, 1978) and bias estimation.…”