*3.5. Discussion of the Experimental Design*

First, I used letters to compose strings for the cognitive load treatment, unlike the conventional use of binary numbers [16]. This design restricts the subjects from using the cognitive load numbers as their inputs for the guessing game. It allows a clear separation of the two tasks, the memorization task and guessing game, and therefore increases the reliability of the treatment effects of cognitive load. I recognized subjects may be able to use other methods to memorize the string of letters, for example, using hand gestures. However, any such method also requires cognitive effort and therefore should not significantly lessen the effects of cognitive load for treatment purposes.

Subjects remained within the fixed pair for the first 16 incentivized guessing games. Since no feedback was given in between games, this design ensures the manipulation of cognitive load being the only source of changing beliefs for any subject. Subjects were different exogenously in terms of cognitive ability, so by staying in the same pair, they carried the same beliefs about their opponents' cognitive abilities throughout the whole session.

Lastly, for each of the 18 incentivized tasks, subjects were given 90 s to make a decision for the guessing game. According to Agranov et al., 90 s is enough for strategic players to make a decision in this type of guessing game [24]. To keep the effect of cognitive load constant across players, I only allowed the subjects to submit their guesses after the 90 s was up. Said practice avoids some subjects naïvely picking a guesses without strategic contemplation for the purpose of achieving correct recalls for the memorization task.
