I have never heard of the randU fallacy. randU actually sounds like a programming term for generating a random number from a uniform distribution. Can you point me to a better explanation of the randU fallacy?
A randU is not part of computing (although computers do have such libraries available).
There are three types of random numbers, randR, randX, and randU. These terms come from mathematics, not computing.
A randR is the same type of random number as see on dice or on a roulette wheel. The random numbers it generates can repeat...even multiple times.
A randN is a non-repeatable number, similar to a deck of cards (once you draw a card, you cannot draw it again until some reset event, such as shuffling the deck).
A randU is a number thought up in someone's head, or an algorithm thought up in someone's head. RandU's in conversation are typically used to embellish some claim, and often wind up as a fallacy in that way.
A paired randR produces the famous Bell Curve.
A fallacy is an error in logic, similar to an arithmetic error in mathematics.
No random number is 'data'.
Random number mathematics is part of the Full Boolean Domain (otherwise known as the Full Discrete Domain), which all machines, including computers, are locked into. In that Domain, the word 'infinity' simply doesn't exist. It has no negative numbers and it has no fractions. Also, division by zero is simply not possible.
They are often translated over to the Real Domain (the mathematics you learned in K-12 education). Because of this, anything that makes use of them loses the natural ability to predict in mathematics. Such branches that use them are probability mathematics and statistical mathematics. Neither branch has the power of prediction inherent in mathematics. Such is the cost of crossing math Domains.
An Argument from randU fallacy is a logical error committed when one attempts to use random numbers of type randU as if they were 'data'.
Random numbers are not about 'distribution'. The moment you try to use 'distribution', you are talking about a randU generator, where the 'distribution' value itself is a randU. Random numbers are not defined by probability or statistical math.
News polls are randU. The questions are unknown, the sample mass is unknown, and a statistical analysis from such numbers it itself a math error simply because it tries to predict (which is not possible), as well as a failure to calculate the margin of error value or to declare and justify the variance. Indeed, the numbers given in the analysis were NOT selected by randN, and NOT normalized by paired randR, as is required.
Now, I know people like to predict elections just as much as predict who will win the football game.
But it's useless and pointless.