PTWhat helps people learn from their own mistakes
Making mistakes at work is unavoidable, and learning from them does not happen by itself. With 176 people who work in organisations, the dissertation measures how much weight lies with the person’s own capabilities, their psychological capital and the way the organisation handles mistakes. Explained in 12 screens.
Mistakes cannot be eliminated, and learning from them depends on many things
Mistakes hurt. Failures stand out and stir up anxiety, anger, shame and guilt, and from school onwards we are rewarded for making fewer mistakes. So the tendency is to hide a mistake rather than study it. The dissertation starts by telling three words apart:
illustration
The two paths summarise the introduction and sections 2.1 and 2.2 (Cannon and Edmondson, 2001; Dahlin, Chuang and Roulet, 2018; Frese and Keith, 2015). The dissertation treats error, failure and mistake as interchangeable, because the boundaries between them blur.
Three groups of factors, each with four parts
Agentic capabilities come from Albert Bandura: they are what lets a person act on their own initiative. Positive psychological capital comes from Fred Luthans: resources such as hope and optimism. Organisational factors describe how the environment handles mistakes. Tap a column:
Height = loading of each part on its group (0 to 1), Figure 5 (capabilities and psychological capital) and Figure 9 (organisational factors). The hollow box goes up to 1. Definitions from sections 1 and 2; statements from the English version of the questionnaire (Appendix B), answered from 1 (strongly disagree) to 7 (strongly agree).
Answer as the participants did
Each person read 57 statements, in random order, and marked from 1 (strongly disagree) to 7 (strongly agree). Here are eight of them, two per block. Answer thinking of your last 12 months at work:
Statements from the questionnaire’s English version (Appendix B). The calculation is just the mean of your answers in each block; it is not a diagnosis and is not compared with the participants (the dissertation does not publish the sample means). Nothing is sent anywhere: it all stays in your browser.
323 people started the questionnaire, and 176 answers remained
The invitation went round messaging apps and social networks, and each participant was asked to pass it on. Anyone who had worked in an organisation in the last 12 months could answer, in Portuguese or English. Beforehand, five professors and 82 people tested the questionnaire, and seven statements were dropped.
Sections 3.1 to 3.6. Convenience sample; the dissertation does not report the profile of the 176 (country, age, sector), only that these questions were used as controls. “Don’t know / prefer not to answer”: up to two were filled with the others’ mean; more than that, and the response was removed.
The dissertation made five bets about learning from mistakes
Learning from mistakes was measured with statements such as “the errors showed me how it is possible to improve my work”. Before seeing the results, bet on each hypothesis. The scoreboard with your bets is on screen 11.
Model from Figure 1 and hypotheses from sections 2.3 to 2.5. Moderation: the environment would change the strength of a link. The profile of the person and of the organisation (country, age, gender, position, sector, size and others) was included as controls.
All three links show up, and together they explain three quarters of learning from mistakes
People who report more agentic capabilities also report more psychological capital and say they learn more from mistakes. The arrows show the strength of each link. Switch the measure or tap an arrow:
Main model, Figure 5 and Table 9, with the 176 responses (PLS-SEM, SmartPLS 4). All with p < 0.001. β = standardised coefficient: how much the other side changes, in standard deviations, when this one rises one standard deviation. f² = how much R² would drop without that arrow.
Agentic capabilities weigh the most, directly and through psychological capital
Reflecting on what you did, managing your emotions, planning and learning by watching colleagues: this set goes along most closely with learning from mistakes. Part of the effect also passes through psychological capital. Move the slider and see what the model predicts:
Table 9. Effect through psychological capital: 0.775 × 0.274 = 0.212; total 0.640 + 0.212 = 0.852, almost equal to the correlation between the two (0.853, Table 8). Our calculations: the dissertation did not test the significance of this indirect effect. The simulator uses the same coefficients, in standard deviations from the mean; it is a reading of the numbers, not a prediction for any one person.
A good environment makes more difference to those with fewer resources of their own
The bet was that an environment that handles mistakes well (a manager and colleagues who help, clear procedures, looking for the cause rather than the culprit) would strengthen the links. The data show something else. Move the environment slider and compare:
If the bet were right
What the data showed
Lines recalculated by us: slope from Tables 12 and 13 (agency 0.593 and interaction −0.073; psychological capital 0.098 and interaction −0.065, both with p < 0.001 and f² 0.042 and 0.041); the height of each line (about 0.18 and 0.16 per standard deviation of the environment) was read from the charts in Figures 8 and 10, because the tables do not report the environment’s own effect. Axes in standard deviations, from −1 to +1, as in the figures. In the left panel, the slopes are made up. With both interactions in the same model, neither was significant (p 0.474 and 0.662), and the dissertation discarded that model for redundancy (Table 10).
With the environment in the model, the weight of psychological capital shrinks
The dissertation ran the model in several ways. The link from agentic capabilities stayed firm in all of them. The one from psychological capital survived the check for the wish to look good, but shrank when the environment came in. Switch the model:
Table 9 (main), Figure 12 (with the 8 social desirability statements), Tables 12 and 13 (with the environment) and Table 14 (controls). Striped = no clear effect (p above 0.05); light = borderline (p = 0.050).
Three bets were confirmed, and the environment acted the opposite way to what was expected
Agentic capabilities are linked to psychological capital (H3) and to learning from mistakes (H2). Psychological capital is too (H1), but that link weakens with the environment in the model. The moderations (H4 and H5) appeared with the opposite sign. Tap a column or a card:
H1 to H3 from Table 9; H4 and H5 from Tables 12 and 13, with the hypothesis numbering of section 2.5 (H4: environment and capabilities; H5: environment and psychological capital). Height = β without its sign; for the moderations, multiplied by 5 to be visible.
Learning from mistakes starts with each person’s capabilities
For the author, organisations gain by developing these capabilities and by replacing punishment of mistakes with conversation about them.
For HR and managers
- Train problem solving, constructive feedback and reflection on practice; offer coaching, mentoring and psychological support. (author)
- Take these capabilities into account when attracting talent. (author)
For the organisation
- Replace the punitive approach with transparency, with regular “lessons learned” sessions and quick feedback. (author)
- Our reading: from screen 9, a good environment matters more to those with less psychological capital or fewer capabilities, such as people who are just starting out.
For people at work
- After a mistake, reflect on what you did and watch how more experienced colleagues handled similar situations. (our reading, based on the questionnaire statements)
What this study does not say
- Who answered. 176 people from a convenience sample, concentrated in Brazil; it does not represent all workers. · according to the author
- Self-assessment. Everything comes from the same questionnaire, about the person themself; it shows what goes together, not what causes what, and does not measure actual mistakes. · author and our observation
- Very close measures. Agentic capabilities and learning from mistakes correlate at 0.853, almost at the limit for being different measures; the author treats them as distinct. · author and our observation
- Translated and shortened scales. Statements that did not work were removed, not rewritten. · according to the author
- What was not tested. The effect that passes through psychological capital had no significance test, and the moderation came out opposite to what was expected. · our observation
Credits
SZPICZKOWSKI, Jairo. Influência das capacidades agênticas e do capital psicológico na aprendizagem a partir do erro nas organizações. Master’s dissertation (Business Administration), Mackenzie Presbyterian University, São Paulo, 2024. Advisor: Prof. Dr. Diógenes de Souza Bido.
Dissertation approved on 2 December 2024, with support from CAPES (Brazil’s federal agency for graduate education). Questionnaire analysed in SmartPLS 4. This version simplifies the work for a general audience.