Executive Function
Belief Update Task
Also called: belief updating task, optimism bias in belief updating, update bias, good news/bad news effect, valence-dependent belief updating
Measure how people revise beliefs about their own future: participants estimate their chance of 32 everyday events, see the average from an official statistic and estimate again, scored as the difference between updates after good and bad news.
10-15 minUntimed
Free to start: 10 free participants per account, then $0.99 per participant for academic research, or any yearly plan. Try it here first, with no account.
- Papers, 2021 to 2025
- 34
- Evidence grade
- C
- Registered clinical studies
- 8
What it measures
Measures how people revise beliefs about their own future after good and bad news. Participants estimate the chance of 32 everyday events, 16 good and 16 bad, happening to them, see the average chance from an official statistic, then estimate each again. The main score is the update bias: how much more estimates move after good news than after bad news.
Constructs
- Belief updating
- Optimism bias
- Learning from feedback
- Risk perception
- Thinking about the future
Research uses
- Judgment and decision-making research
- Optimism and mood research
- Risk communication research
- Within-study group comparison
How the paradigm works
The Belief Update Task measures how people revise beliefs about their own future when they learn how common an event is. The participant first estimates the chance that each of 32 everyday events will happen to them, moving a slider from 0 to 100%. After each estimate the average chance appears, a figure from an official statistic shown with its population, place and year. A second part asks for every estimate again, in a new order and without the averages, and a last part (standard; a setting leaves it out) asks for each average shown, the memory check of the original design. Half the events are good, such as a pay rise or a trip of four nights or more, and half are bad, such as an online order arriving late or a scam message; none is about illness, injury, death, a crime against a person or weight. The events are written for this task in each language, and the averages come from Eurostat, the Federal Reserve Board's Survey of Household Economics and Decisionmaking, the Office of Rail and Road, the Department for Culture, Media and Sport and the Department for Transport in the United Kingdom, and the Statistics Bureau of Japan. One practice event, a fair coin, comes first. There is no time limit and no feedback on the estimates.
Scores
- Update Bias%
- Mean update after good news minus mean update after bad news, in percentage points. An update is how far the second estimate moved toward the average.
- Above 0, estimates moved more after good news than after bad news
- Update After Good News%
- Mean move of the second estimate toward the average, in percentage points, for events whose average was better news than the first estimate.
- Informational
- Update After Bad News%
- Mean move of the second estimate toward the average, in percentage points, for events whose average was worse news than the first estimate.
- Informational
- Learning Rate Biasscore
- Mean learning rate after good news minus mean learning rate after bad news. A learning rate is the update divided by the distance between the first estimate and the average.
- Above 0, good news moved estimates by a larger share of the gap
- Learning Rate After Good Newsscore
- Mean of each update divided by the distance between the first estimate and the average, for good news; 1 means the second estimate reached the average.
- Informational
- Learning Rate After Bad Newsscore
- Mean of each update divided by the distance between the first estimate and the average, for bad news.
- Informational
- Good-News Eventscount
- Events with both estimates whose average was lower than the first estimate for a bad event, or higher for a good event.
- Fewer than 4 is flagged
- Bad-News Eventscount
- Events with both estimates whose average was higher than the first estimate for a bad event, or lower for a good event.
- Fewer than 4 is flagged
- No-News Eventscount
- Events whose first estimate equaled the average; they are left out of the update scores.
- Informational
- First-Estimate Gap, Good News%
- Mean distance between the first estimate and the average for good-news events, in percentage points.
- Informational; compare with the bad-news gap
- First-Estimate Gap, Bad News%
- Mean distance between the first estimate and the average for bad-news events, in percentage points.
- Informational; compare with the good-news gap
- Memory Error, Good News%
- Mean distance between the average recalled at the end and the average shown, for good-news events; empty when recall is left out.
- Close to the bad-news value means the bias does not come from memory
- Memory Error, Bad News%
- Mean distance between the average recalled at the end and the average shown, for bad-news events; empty when recall is left out.
- A mean of the two above 20 points is flagged
- Unchanged Estimates%
- Share of events whose second estimate equaled the first.
- 1 is flagged
- Median Estimate Timems
- Median time from an estimate's screen appearing to its confirmation, over first and second estimates of at least 100 ms.
- Informational; the task has no time limit
- Answers Under 100 mscount
- Answers confirmed less than 100 ms after their screen appeared.
- 0 expected
- Events Answeredcount
- Events with both a first and a second estimate in the scored run.
- Fewer than half the planned events invalidates
- Averages Recalledcount
- Averages recalled in the last part; empty when recall is left out.
- Informational
- Planned Eventscount
- Events the run planned: 32, 16 good and 16 bad.
- Informational
Use and evidence
34 papers from 2021 to 2025 name this task in their title or abstract. That ranks it 532 of 711 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 9 |
| 2024 | 3 |
| 2023 | 5 |
| 2022 | 7 |
| 2021 | 10 |
| 2020 | 4 |
| 2019 | 2 |
| 2018 | 2 |
| 2017 | 1 |
| 2016 | 0 |
| 2015 | 1 |
| 2014 | 1 |
| 2013 | 0 |
| 2012 | 1 |
| 2011 | 0 |
| 2010 | 0 |
| 2009 | 0 |
| 2008 | 0 |
| 2007 | 0 |
| 2006 | 0 |
| 2005 | 2 |
| 2004 | 0 |
| 2003 | 1 |
| 2002 | 1 |
| 2001 | 0 |
| 2000 | 0 |
Use grade D
30 to 99 papers from 2021 to 2025.
Evidence grade C
At least 2 test-retest papers, 30 reliability or validity papers or 3 registered clinical studies.
- Papers on test-retest reliability
- 0
- Papers on reliability or validity
- 2
- Registered clinical studies using it as an outcome
- 8
Measured on ConductCognition
Pooled across the studies that feed our platform norms, by device group. A figure needs 50 or more people.
No pooled figure yet.
How the grades work
Use: A is 1,000 or more papers from 2021 to 2025, B 300 to 999, C 100 to 299, D 30 to 99 and E fewer than 30. Evidence: A means at least 25 papers report its test-retest reliability and at least 25 registered clinical studies use it as an outcome; B, at least 8 test-retest papers, and either 5 registered clinical studies or 150 reliability or validity papers; C, at least 2 test-retest papers, 30 reliability or validity papers or 3 registered clinical studies; D, little published measurement evidence found. The counts measure use. They include every paper that names the task in its title or abstract. Grades describe the published research on this task, not ConductCognition's version of it.
Source: ConductCognition test census 0.58.0, counted 31 October 2026 from OpenAlex and ClinicalTrials.gov.
Task design
The trial sequence as the test runs in ConductCognition.
- Events
- 32 everyday events written for this library in each language, 16 good and 16 bad, such as a pay rise, a trip to the cinema, a parcel arriving late or a scam message. None is about illness, injury, death, a crime against a person or weight.
- Averages
- Each average is a published figure from an official statistic, rounded to a whole percent and shown with its population, place and year, from 10% to 85%. Sources: Eurostat, the Federal Reserve Board's Survey of Household Economics and Decisionmaking, UK statistics on rail, buses and cultural life, and the Statistics Bureau of Japan.
- Answer
- A slider from 0 to 100% that starts with no value chosen, for the chance that the event will happen to the participant.
- Parts
- First estimates, each followed by its average; every estimate again, without the averages; then each average recalled. Each part has its own order drawn from the session's seed. One practice event, a fair coin, comes first; there is no time limit.
- Setting
- Recall of the averages: asked at the end (standard) or left out, which shortens the session by about 3 minutes.
- Scores
- The update bias, the mean update after good news minus the mean after bad news (the main score); the same comparison with learning rates; the number of good-news, bad-news and no-news events; the first estimates' distance from the averages; and memory errors for the averages.
Each average names its population, place and year on screen, so participants see what the figure describes. The scores describe how estimates change after good and bad news; they support research and do not judge a person.
Quality checks
Each session is marked valid, needs review or invalid, with the reason, before you see its scores.
Scored rows: belief_update_first from the run with the most of them, latest on ties, and that run's belief_update_second and belief_update_recall rows. An estimate of one's own chance has no right answer, so estimates far from the averages never invalidate. Fewer than half the planned events with both estimates invalidates, and any other incomplete run is flagged. Fewer than 4 good-news or bad-news events, no second estimate that differs from its first, and a mean memory error above 20 percentage points are flagged for review. Answers under 100 ms are counted as anticipations, and answers under 200 ms on at least half the answers invalidate. A hidden tab is flagged.
Why these rules
There is no correct option to compare with chance, so accuracy is not judged against chance. How far estimates move toward the averages is what the task measures, so not moving them is a result, not a failed task; it is flagged when no estimate moves at all, and a poor memory for the averages is flagged because the update scores assume they were read (Sharot, Korn and Dolan 2011). An answer under 200 ms is faster than a simple visual reaction, so it cannot reflect the item. Following response-time effort screening for untimed tests (Wise and Kong 2005), the session is invalid only when at least half the answers are that fast. The averages are published figures, so a hidden tab could allow a search for them; it is flagged.
No percentile. Scores and quality flags only.
Running this test
Settings you can change
| Setting | Standard | Allowed |
|---|---|---|
| Recall of the averages | At the end (standard) | At the end (standard), Left out |
Yearly plans can change settings within these limits. Free and pay-as-you-go studies use the standard settings.
Need a setting this test does not have?
Devices and languages
- Untimed test: runs on desktops, laptops, tablets and phones. Participant devices
- Languages: English, Spanish, French, German, Italian, Portuguese (Brazil), Chinese (Simplified) and Japanese.
- Available on: Every plan, including the free plan.
- No percentile. Scores and quality flags only. How norms work
- Recruiting online? See the setup guides for Prolific, Sona Systems, CloudResearch Connect and Qualtrics.
Cite and describe
Participants completed the Belief Update Task in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Sharot T, Korn CW, Dolan RJ, 2011; Garrett N, Sharot T, 2014), with the standard settings (recall of the averages: At the end). The main score was update bias (mean update after good news minus mean update after bad news, in percentage points).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Sharot, T., Korn, C. W., & Dolan, R. J. (2011). How unrealistic optimism is maintained in the face of reality. Nature Neuroscience, 14(11), 1475-1479. https://doi.org/10.1038/nn.2949The method: an estimate of one's own chance of each adverse event, the average chance, then a second estimate, with a memory check of the averages. Estimates moved more after good news than after bad news.
- Garrett, N., & Sharot, T. (2014). How robust is the optimistic update bias for estimating self-risk and population base rates? PLoS ONE, 9(6), e98848. https://doi.org/10.1371/journal.pone.0098848The asymmetry between updates after good and bad news found again in estimates of one's own risk.
- Shah, P., Harris, A. J. L., Bird, G., Catmur, C., & Hahn, U. (2016). A pessimistic view of optimistic belief updating. Cognitive Psychology, 90, 71-127. https://doi.org/10.1016/j.cogpsych.2016.05.004A critique arguing that the asymmetry could come from how the method measures updating.
- Garrett, N., & Sharot, T. (2017). Optimistic update bias holds firm: Three tests of robustness following Shah et al. Consciousness and Cognition, 50, 12-22. https://doi.org/10.1016/j.concog.2016.10.013A reply that found the asymmetry for positive life events after correcting confounds the authors identified in the critique's experiments.
- Kuzmanovic, B., & Rigoux, L. (2017). Valence-dependent belief updating: Computational validation. Frontiers in Psychology, 8, 1087. https://doi.org/10.3389/fpsyg.2017.01087Computational models of the updates found higher learning rates for good news than for bad news.
Source papers
- Sharot T, Korn CW, Dolan RJ (2011). How unrealistic optimism is maintained in the face of reality. Nature Neuroscience, 14(11):1475-1479.Paradigm source
- No normative data for this task.Paradigm source, N/A
Questions
How long does the Belief Update Task take?
About 10 to 15 minutes, as listed in the test library.
What device does a participant need?
Untimed test: runs on desktops, laptops, tablets and phones. The browser must be Chrome, Edge, Firefox or Safari.
What scores does it produce?
Update Bias, Update After Good News, Update After Bad News, Learning Rate Bias, Learning Rate After Good News, Learning Rate After Bad News, Good-News Events, Bad-News Events, No-News Events, First-Estimate Gap, Good News, First-Estimate Gap, Bad News, Memory Error, Good News, Memory Error, Bad News, Unchanged Estimates, Median Estimate Time, Answers Under 100 ms, Events Answered, Averages Recalled and Planned Events. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Belief Update Task used in research?
34 papers from 2021 to 2025 name it in their title or abstract. That ranks it 532 of 711 test families in the ConductCognition test census (use grade D). Source: ConductCognition test census 0.58.0, counted 31 October 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Belief Update Task have?
In the published research, 0 papers report its test-retest reliability and 8 registered clinical studies use it as an outcome (evidence grade C in the ConductCognition test census).
Does it report a percentile?
No. Scores and quality flags only.
Which settings can a researcher change?
Recall of the averages (At the end (standard), Left out). Settings can be changed on yearly plans; free and pay-as-you-go studies use the standard settings.
Which languages can participants use?
Participants can take this test in English, Spanish, French, German, Italian, Portuguese (Brazil), Chinese (Simplified) and Japanese, with its words in the same language. English is on every plan, and the other languages come with yearly plans. Have a native speaker read the machine-translated text before launch.
Is this a clinical test?
No. ConductCognition is for research use only. It runs, scores and exports the test; clinical interpretation is outside the platform.
Every test: the full test list.
Run this test in your next study
Start free with 10 free participants on any test, then use participant credits or a yearly plan.