**Hypotheses and expectations** While corruption often leads to direct harm to society or some other third party, sometimes the harm is uncertain. For example, the decision which is supposed to be influenced by a bribe may not depend only on one employee. This may be the case when the decision is made by a committee or when the decision has to be approved by multiple people. It is possible that people will be more likely to take a bribe in this case because responsibility for the decision is shared among multiple decision makers (Darley & Latané, 1968; Feldman & Rosen, 1978). We will study the effect of uncertainty of consequences of taking a bribe by manipulating the probability that the charity loses money after a participant performs the task incorrectly. While in the default variant of the task charity loses money whenever a participant does not sort an object according to its color, the experiment will test the effect of the probability of the loss while holding the expected value of the loss constant. **Methods** *Participants* Participants will be recruited from a laboratory subject pool to participate in an on-line study. The planned sample size is 300, but the final sample size might slightly differ if multiple people start the experiment at the same time. The experiment will have sufficient power (.80) to detect an effect *d* = 0.23. A more precise power analysis is complicated by the nature of the repeated measures design of the experiment and by complexity of mixed-effect regression, which will be used for analysis. We will exclude participants with missing data from 3 or more trials, which may occur, for example, in case of a repeated internet connection failure. The analysis will be conducted only with data from participants who will finish the whole study (i.e., get to the final screen of the application). We will also exclude participants who will at least 10 times sort the object according to neither its color, nor its shape, suggesting random responding. Participants are also informed that the task is terminated if they sort the object incorrectly four times in a row. Participants with a task terminated in this manner will be also excluded from analysis. *Materials and procedure* The experiment will use a task simulating routine work during which the worker is sometimes given an opportunity to take a bribe (Vranka & Bahník, 2018). In particular, the task is aimed to simulate work of a bureaucrat or a public employee in general. As a part of the task, participants are told to sort objects running on a computer screen according to their color. The objects may have three possible shapes and colors. The sorting is done by pressing one of three keys, each of which is associated with a single color and shape. Shapes and colors associated with the three keys are randomly determined for each trial and they are displayed prominently at the bottom of the screen. At the beginning of the task, 2000 points (corresponding to 200 CZK, ~8 USD) are allotted to a charity. If a key response leads to a wrong assignment to a color, the charity loses a certain amount of points that differs by experimental condition. Participants will be divided in one of three groups. In the control group, the charity will lose 200 points whenever a participant sorts the object incorrectly. In the medium probability group, the charity will lose 400 points with a 50% probability when a participant sorts the object incorrectly. In the low-probability group, the loss will be 2000 points and the probability will be 10%. The loss—if it occurs—is highlighted by increasing the size of the font and its color to red of the text displaying the current reward for the charity for one second. The loss simulates negative societal effects of not performing given work correctly. The task is also automatically terminated if the object is sorted incorrectly four times in a row. Participants are informed about this rule and the number of previous incorrect responses is displayed on the screen with the corresponding number of red X’s. Participants get a fixed reward of 3 points for each sorted object, which represents the salary given to a worker for performing his or her job. Finally, in trials where the two sorting criteria are mismatched, there is a 22.5% probability that a given object will be associated with a “bribe”. These objects are shown with a number corresponding to the value of the bribe, which a participant gets if he or she sorts the object according to its shape. The bribe size will vary from 30 to 180 points in increments of 30 and it is determined randomly for each trial with an associated bribe. Each participant will be given 200 trials of the task; i.e., 200 objects to sort. The number of the trial as well as the money earned for oneself and the charity will be displayed on the screen during the whole task. The responses will determine participants’ reward and money gained (or lost) for the charity. The experiment will be conducted on-line using a custom-written Python (Django) and Javascript web application running on www.pythonanywhere.com. Participants will be explained the task, complete 10 practice trials and then proceed with the task itself. The points participants earn during the task will be converted to a monetary reward using the conversion rate 10 points = 1 CZK. **Analysis** The script that will be used for analysis can be found in Files. *Bribe taking* Trial-level analysis will be conducted using mixed-effect linear regression. The correctness of object classification will serve as the dependent variable. The trials incorrectly sorted according to both shape and color as well as trials without a bribe will be excluded. Random intercepts for participants will be included. Order of the trial, squared order of the trial (rescaled to range from -0.5 to 0.5), and linear and quadratic contrasts for bribe size will be included as covariates. The groups will be compared using linear and quadratic contrasts. Random slopes for participants will be included for the four covariates. Correlations will be included only between the random effects for the same variable (order of the trial and squared order of the trial; linear and quadratic contrasts for bribe size). If the model suffers from convergence issues, random slopes or correlations between random effects will be removed to improve the model fit. **References** - Darley, J. M., & Latané, B. (1968). Bystander intervention in emergencies: Diffusion of responsibility. *Journal of Personality and Social Psychology, 8*, 377-383. - Feldman, R. S., & Rosen, F. P. (1978). Diffusion of responsibility in crime, punishment, and other adversity. *Law and Human Behavior, 2*, 313-322. - Vranka, M. A., & Bahník, Š. (2018). Predictors of bribe-taking: the role of bribe size and personality. *Frontiers in Psychology, 9*, 1511.