Decision Making
Ellsberg Urns Task
Also called: Ellsberg paradox task, ambiguity aversion task, two-color Ellsberg urn, known and unknown urns, matching probability
Measure aversion to unknown odds: participants bet on a ball of their color from a box with a known mix or one with an unknown mix, scored as ambiguity aversion from the matching probability.
4-6 minUntimed
Free to start: 10 free participants per account, then $0.99 per participant for academic research, or any yearly plan. Preview any test free after you sign up.
- Papers, 2021 to 2025
- 63
- Evidence grade
- D
- Registered clinical studies
- 1
What it measures
Measures aversion to unknown odds. The participant bets on a ball of their own color from one of two boxes of 100 balls: one with a known mix, from 10 to 90 balls of that color, and one whose mix is not shown. The switch point gives a matching probability, and ambiguity aversion is 0.5 minus it.
Constructs
- Ambiguity aversion
- Decision under uncertainty
- Risk and ambiguity
- Economic preferences
Research uses
- Behavioral economics and neuroeconomics research
- Decision-making under uncertainty
- Finance and investment behavior
- Within-study group comparison
How the paradigm works
The Ellsberg Urns Task measures aversion to ambiguity: preferring a bet whose odds are known to one whose odds are not (Ellsberg, 1961). The participant chooses their own color, blue or orange, before any instructions, so the unknown box cannot be set against it (Kocher, Lahno and Trautmann, 2018). Each choice then shows two boxes of 100 balls: one whose mix is shown, with 10 to 90 balls of the participant's color, and one whose mix is not, the same box in every choice. The participant picks the box a ball would be drawn from, and a ball of their color wins points. The first choice offers the even box, 50 of 100, against the unknown one, Ellsberg's own question; then every known box comes once in each of 3 rounds (1 to 4 as a setting), in an order drawn from the session's seed. Nothing is drawn during the task. At the end one choice is drawn and a ball is drawn from the box chosen in it; a ball of the participant's color wins 100 points. One check question with feedback, on the chance of drawing one's color from a known box, comes first, and there is no time limit. The study decides what the points are worth, and the payment line says whether they are exchanged for money.
More about the paradigm
The main score is ambiguity aversion: 0.5 minus the matching probability, the known chance at which a person switches from the unknown box to the known one (Dimmock, Kouwenberg and Wakker, 2016). Above 0 prefers known odds, below 0 prefers unknown odds, and 0 treats them alike. The matching probability is read from the share of choices of the unknown box at each known box; for a person who switches once, it is the midpoint between the two boxes around the switch, as in Kocher and colleagues. The share of choices of the known box at each number of balls traces the whole choice curve, and the first choice is the classic Ellsberg choice on its own. In a representative Dutch sample paid for real, the mean matching probability was 0.40, an ambiguity aversion of 0.10, and 68% chose the known even box at first (Dimmock and colleagues, 2016). Choices under 100 ms are counted as anticipations and left out. Scores support research and do not diagnose.
Scores
- Ambiguity Aversionscore
- 0.5 minus the matching probability.
- Above 0 prefers known odds, below 0 unknown odds
- Matching Probability%
- The known chance, from 0.05 to 0.95, at which choices switch from the unknown box to the known box: the lowest cell edge plus each known box's cell width times its share of unknown-box choices.
- 0.5 treats known and unknown odds alike
- First Choice Known0/1
- 1 when the first choice, the even known box against the unknown box, took the known box.
- 1 is the classic Ellsberg choice
- Known Box at 10%
- Share of choices of the known box when it held 10 balls of the participant's color.
- Informational
- Known Box at 20%
- Share of choices of the known box when it held 20 balls of the participant's color.
- Informational
- Known Box at 30%
- Share of choices of the known box when it held 30 balls of the participant's color.
- Informational
- Known Box at 40%
- Share of choices of the known box when it held 40 balls of the participant's color.
- Informational
- Known Box at 45%
- Share of choices of the known box when it held 45 balls of the participant's color.
- Informational
- Known Box at 50%
- Share of choices of the known box when it held 50 balls of the participant's color, the even box.
- Above 0.5 prefers known odds
- Known Box at 55%
- Share of choices of the known box when it held 55 balls of the participant's color.
- Informational
- Known Box at 60%
- Share of choices of the known box when it held 60 balls of the participant's color.
- Informational
- Known Box at 70%
- Share of choices of the known box when it held 70 balls of the participant's color.
- Informational
- Known Box at 80%
- Share of choices of the known box when it held 80 balls of the participant's color.
- Informational
- Known Box at 90%
- Share of choices of the known box when it held 90 balls of the participant's color.
- Informational
- Choice Consistency%
- Share of choices that agree with the best single switch from the unknown box to the known box.
- Below 0.75 is flagged for review
- Check Correctflag
- 1 when the check question about the chance of a known box was answered correctly, 0 when not.
- 1 supports interpretation
- Median Decision Timems
- Median time from a choice's screen to the choice.
- Informational
- Anticipationscount
- Choices made under 100 ms after the boxes appeared, left out of the scores.
- 0 expected
- Choices Answeredtrials
- Choices made, out of 33 as standard.
- Informational
- Points Wonpoints
- Points from the drawn choice: 100 when the ball drawn was the participant's color, 0 when not.
- Informational
Use and evidence
63 papers from 2021 to 2025 name this task in their title or abstract, down 16% compared with 2016 to 2020. That ranks it 235 of 395 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 11 |
| 2024 | 8 |
| 2023 | 11 |
| 2022 | 13 |
| 2021 | 20 |
| 2020 | 9 |
| 2019 | 10 |
| 2018 | 19 |
| 2017 | 23 |
| 2016 | 14 |
| 2015 | 22 |
| 2014 | 14 |
| 2013 | 7 |
| 2012 | 16 |
| 2011 | 17 |
| 2010 | 8 |
| 2009 | 14 |
| 2008 | 14 |
| 2007 | 12 |
| 2006 | 8 |
| 2005 | 9 |
| 2004 | 4 |
| 2003 | 6 |
| 2002 | 5 |
| 2001 | 7 |
| 2000 | 2 |
Use grade D
30 to 99 papers from 2021 to 2025.
Evidence grade D
Little published measurement evidence found.
- Papers on test-retest reliability
- 0
- Papers on reliability or validity
- 15
- Registered clinical studies using it as an outcome
- 1
Note on the counts: includes decision-theory papers that run no task.
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 0.14.0, counted 2 November 2026 from OpenAlex and ClinicalTrials.gov.
Task design
The trial sequence as the test runs in ConductCognition.
- Color
- The participant chooses blue or orange before any instructions, so the unknown box cannot be set against that color. The color stays for the whole task.
- Boxes
- Two boxes of 100 balls on each screen. The known box shows 10, 20, 30, 40, 45, 50, 55, 60, 70, 80 or 90 balls of the participant's color. The unknown box is the same in every choice; its mix, drawn from the session's seed, is not shown.
- Choices
- The even box (50) first, then every known box once in each of 3 rounds (standard; 1 to 4), in an order drawn from the session's seed: 33 choices. Click or tap a box; no time limit, and nothing is drawn during the task.
- Check
- One question with feedback before the choices: the chance of drawing one's color from a box with 35 of 100 balls of that color.
- Points
- One choice drawn from the session's seed is played at the end: a ball of the participant's color from the box chosen in it wins 100 points. The study decides what points are worth, and the payment line says whether they are exchanged for money.
- Scores
- Ambiguity aversion, 0.5 minus the matching probability: the lowest cell edge, 0.05, plus each known box's cell, halfway to its neighbors, times its share of unknown-box choices. Also the first choice, the share of known-box choices at each box and how well the choices fit one switch. Choices under 100 ms are left out as anticipations.
- Checks
- Fewer than half the choices made is invalid. Under 75% of choices agreeing with a single switch and a wrong check answer are flagged and never exclude.
Ellsberg (1961) posed one bet on two urns. This test follows the modern elicitation, a list of known chances against one ambiguous box with the winning color chosen first (Dimmock and colleagues, 2016; Kocher and colleagues, 2018), here as separate choices in a seeded order, repeated over rounds. It measures the attitude to one likely event; insensitivity to likelihood needs boxes of ten colors, which this test does not offer. Matching probabilities depend on the stakes and on whether choices are paid, so compare results within a study. The scores support research and do not diagnose.
Quality checks
Each session is marked valid, needs review or invalid, with the reason, before you see its scores.
Scored rows: the scored run's choices between the known box and the unknown box, 11 in each round. No box is correct, so accuracy is not judged against chance, and choosing one box throughout does not invalidate. Fewer than half the planned choices answered is invalid, and any other incomplete run is flagged. No estimate of ambiguity aversion, under 75% of choices agreeing with a single switch point and a wrong check answer are flagged for review and never exclude. Choices under 100 ms are counted as anticipations and left out, and choosing in under 200 ms on at least half the choices invalidates.
Why these rules
There is no correct option to compare with chance, so accuracy is not judged against chance. Preferring a box with known odds, or one with unknown odds, is the attitude the test measures, not an error. A person with one matching probability switches once from the unknown box to the known box as its chance rises, so choices that rarely agree with any single switch look like choosing at random; real but noisy preferences can do that too, so this flags and never excludes. 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.
No percentile. Scores and quality flags only.
Running this test
Settings you can change
| Setting | Standard | Allowed |
|---|---|---|
| Rounds | 3 | 1 to 4 |
| Payment line | Hypothetical points (standard) | Hypothetical points (standard), The study pays for points |
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 Ellsberg Urns Task in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Ellsberg D, 1961), with the standard settings (rounds: 3; payment line: Hypothetical points). The main score was ambiguity aversion (0.5 minus the matching probability).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Ellsberg, D. (1961). Risk, ambiguity, and the Savage axioms. The Quarterly Journal of Economics, 75(4), 643-669. https://doi.org/10.2307/1884324
- Dimmock, S. G., Kouwenberg, R., & Wakker, P. P. (2016). Ambiguity attitudes in a large representative sample. Management Science, 62(5), 1363-1380. https://doi.org/10.1287/mnsc.2015.2198Matching probabilities and the ambiguity aversion index.
- Kocher, M. G., Lahno, A. M., & Trautmann, S. T. (2018). Ambiguity aversion is not universal. European Economic Review, 101, 268-283. https://doi.org/10.1016/j.euroecorev.2017.09.016The winning color chosen first, the even choice and the list of known chances.
- Trautmann, S. T., & van de Kuilen, G. (2015). Ambiguity attitudes. In G. Keren & G. Wu (Eds.), The Wiley Blackwell Handbook of Judgment and Decision Making (pp. 89-116). Wiley. https://doi.org/10.1002/9781118468333.ch3A survey of how ambiguity attitudes are measured and what experiments found.
Source papers
- Dimmock, Kouwenberg and Wakker (2016): in a representative Dutch sample paid for real (666 people), the mean matching probability for the even box was 0.40 (SD 0.24), an ambiguity aversion of 0.10, and 68% chose the known even box at first. Kocher, Lahno and Trautmann (2018): ambiguity aversion for likely gains, but not for losses or unlikely gains.Paradigm source, Adults
Questions
How long does the Ellsberg Urns Task take?
About 4 to 6 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?
Ambiguity Aversion, Matching Probability, First Choice Known, Known Box at 10, Known Box at 20, Known Box at 30, Known Box at 40, Known Box at 45, Known Box at 50, Known Box at 55, Known Box at 60, Known Box at 70, Known Box at 80, Known Box at 90, Choice Consistency, Check Correct, Median Decision Time, Anticipations, Choices Answered and Points Won. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Ellsberg Urns Task used in research?
63 papers from 2021 to 2025 name it in their title or abstract, down 16% compared with 2016 to 2020. That ranks it 235 of 395 test families in the ConductCognition test census (use grade D). Source: ConductCognition test census 0.14.0, counted 2 November 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Ellsberg Urns Task have?
In the published research, 0 papers report its test-retest reliability and 1 registered clinical study uses it as an outcome (evidence grade D in the ConductCognition test census).
Does it report a percentile?
No. Scores and quality flags only.
Which settings can a researcher change?
Rounds (1 to 4) and payment line (Hypothetical points (standard), The study pays for points). 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. 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.