Recency User Input
Upon running the execution with the same character in mind, we discovered that the Akinator takes previous user inputs and answer habits into consideration while traversing its network of nodes. For three back-to-back iterations, we outputted popular quarterback "Aaron Rodgers" three times, answering each question the same for each iteration. During the first iteration, Akinator started with its initial "Is your character fictional?" question. The second iteration yielded the initial question of "Is your character male?" almost assuming the user was going to pick a non-fictional character based on the output of the first iteration, leading to a successful result quicker. The third consecutive iteration, produced the initial question "Is your character an athlete?", as if it knew our output habits.
With this experiment, we conclude that the neural network is structured in a way that previous consecutive user inputs are considered during traversal and especially with the initial starting node itself. The network assumes that characters from the same user are typically all located in a smaller sub-group than the set of all possibilites.
Upon running the execution with the same character in mind, we discovered that the Akinator takes previous user inputs and answer habits into consideration while traversing its network of nodes. For three back-to-back iterations, we outputted popular quarterback "Aaron Rodgers" three times, answering each question the same for each iteration. During the first iteration, Akinator started with its initial "Is your character fictional?" question. The second iteration yielded the initial question of "Is your character male?" almost assuming the user was going to pick a non-fictional character based on the output of the first iteration, leading to a successful result quicker. The third consecutive iteration, produced the initial question "Is your character an athlete?", as if it knew our output habits.
With this experiment, we conclude that the neural network is structured in a way that previous consecutive user inputs are considered during traversal and especially with the initial starting node itself. The network assumes that characters from the same user are typically all located in a smaller sub-group than the set of all possibilites.