To do so there are three different ways.
The first and easiest would be using a random trigger, which saves a value between 1 and the amount of filter paths used in a user defined variable.
This value can then be used for filtering purposes.
The disadvantage of this method ist, that the uniform distribution is only being strived for, but won't be 100% accurate. The higher the number of participants finishing your survey is, the more accurate it gets, but especially for low participant numbers this might lead to bigger differences in the distribution.
Another possibility would be using quotas. For that you need to set up as many quotas as you have filter paths und choose "Matching, open quota with the fewest participants allocated to it" under "Allocation for all quotas".
Please make sure, that the quota cannot be full and the filling condition is always true (i.E. v_1 != 12345) and execute the quotation at a fitting position in your survey.
You can then filter for the quota ID in your questionnaire.
This, just like the last method ensures an accurate uniform distribution.
The last option would be to utilize the running number. This would use the modulo function.
Here, a value n gets divided by another value m and the remaining rest of this division is being stored.
If you want to use this to ensure uniform distribution, you divide the LFDN by the number of relevant filter paths. For two paths that would be:
For even LFDN's this would give out the rest 0, for uneven numbers it would be 1. This calculation has to be drawn in a recoding trigger that saves the result into a user defined variable.
In the survey you can then filter for c_0001 = 0 and c_0001 = 1 and so on.