Overconfidence in Self-Efficacy
In cognitive and educational psychology, overconfidence is a phenomenon that emerges in self-assessment situations such as one that requires students to measure self-efficacy.
Many research studies have discovered that humans tend to recall positive personality traits more readily over negative ones, evaluate themselves more positive over others, overestimate their abilities and overestimate the depth of their understanding.
This article is about the effect of overconfidence on learning and motivational constructs such as performance and metacognition.
Contents
Definition
Overconfidence, in an educational psychology setting, is a phenomenon in which a person’s confidence (typically self-efficacy) in his or her performance on a task is significantly (statistically) higher than his or her actual task performance. Many studies done on overconfidence have used the same ranking or scoring construct for both measuring confidence magnitude and evaluating actual performance. For instance, in a study based on the game Mastermind, the participants rated their self-efficacy in the form of numbers of trials required before reaching correct solutions, which is also how performance is generally evaluated in the game.[cite Vancouver]
Overconfidence in Self-Efficacy and Motivational Constructs
In the Social Cognitive Theory, behavioral, cognitive and environmental constructs interact with each other reciprocally. Thus, under such view it is generally accepted that self-efficacy is a strong, positive predicator of performance. [cite bandura] Even then, relatively little is known about the effect self-efficacy judgments have on learning and motivational constructs like performance, goal orientation and affect.
Overconfidence in Self-Efficacy
There has been a considerable amount of research that lent support to the predictive and mediational role of self-efficacy in learning [cite MAddux and other stuff from Pajares]; however, there has also been some correlational studies that exhibited non-positive correlations between self-efficacy and performance. For example, in a study done on mathematics performances of eighth grade gifted and regular education students, Frank Pajares discovered that even though gifted girls surpassed gifted boys in performance, the two groups did not differ in self-efficacy. In addition, even though most students observed were overconfident, gifted students tend to be better calibrated (or overconfident by a smaller margin) than regular education students. The correlations shown in this study alone seemed to have revealed a more complicated picture of self-efficacy than one that strictly correlates positively with performance.
Some research studies have pointed out that overconfidence in self-efficacy occurs more readily in cognitive complex tasks [cite stone], especially in ones involving causal explanations [cite keil].
Negative Effects of High Self-Efficacy
Even though high self-efficacy is thought to be a positive predictor of performance, a number of papers have documented otherwise.
Overconfidence and Effort
In a study conducted by Stone, when overconfidence is induced (by telling participants that the system automatically adjusts their choice accuracy to increase their performance score so if they work hard they are expected to outperform 90% of the participants), instead of exhibiting corresponding increases in effort, attention to strategy or performance, the participants spent less time on task and performed noticeably worse than those induced with mildly negative expectations (by telling them no choices were adjusted so they need to work hard in order to outperform 50% of the participants). These results suggest that mildly negative self-efficacy beliefs may have more profound motivating effects than a strictly positive and a strongly negative one.
Caveats and Further Research
Stone’s papers employed longer time on task for participants with mildly negative self-efficacy beliefs as evidence for increase in effort and increase in attention to strategy. There is more to learned about what strategies participants employed and if differences in self-efficacy beliefs predict differences in metacognitive strategies.
Overconfidence and Performance
In a more recently study, Vancouver et al. (2001), in addition to reconfirming a positive correlation between self-efficacy and performance in an analysis between persons, also discovered that performance is a positive predictor of subsequent self-efficacy, and self-efficacy as a negative predicator of subsequent performance in a with-in person analysis done across time.
In a study done on the game Mastermind[cite Vancouver], results indicated that participants who were induced with overly positive self-efficacy beliefs (by automatically configuring the first few Mastermind puzzles to match their guesses, creating illusions that the participants are very good at the game), estimated their performance to be an average of 5.40 (SD = 1.66) trials before reaching a solution, compared to a control group (self-efficacy not manipulated) with an average of 6.61 (SD = 1.58) trials. In reality, the experimental group’s performance averaged at 7.10 (SD = 1.10) trials and the control group at 7.17 (SD = 1.10) with a t-test p value of 0.784. The difference in actual performance, Vancouver et al. argued, is not statistically significant.
Vancouver et al. further argued that the positive correlation between self-efficacy and performance at a-between-person level of analysis may be due to individuals’ differences in performance caused by actual differences in capacity to organize and execute abilities that also influences individual differences in beliefs in capacities.[cite Vancouver]
Caveats and Further Research
As noted above, some of the data collected by Vancouver et al. did not reflect the hypothesis that self-efficacy is a negative predictor of performance. Even though Vancouver et al. applied hierarchical linear models to produce analyses that supposedly confirm the hypothesis, the fact that their data’s contrary pattern was so close to passing a test of statistical significance (p = 0.784) is quite a worry. It may be desirable to design an experiment to reconfirm Vancouver et al.’s hypothesis with less ambiguity.
In addition, choices of experiment methods in both Vancouver et al. (2001, 2002) and in Stone (1994) hardly involve mastery goals; thus there is a reasonable possibility that participants, when satisfied with expected performance (overconfidence) or completely despaired (severe underconfidence), have no other incentive to employ more sophisticated metacognitive strategies. It may be interesting to test similar protocols on activities involving mastery goals to see if overconfidence and underconfidence boasts similar influences.
References
- [1] Pajares
- [2] Stone
- [3] Vancouver
- [4] Keil
