Monster game: Which one is the real monster? From information search to education tools

Autor:Björn Meder, Jonathan Nelson, Heike Knauber, Hannes Helmut Nepper, Laura Martignon
Institution:Lange Nacht der Wissenschaften Berlin + Potsdam
Veröffentlicht: 13.06.2015

How do children and adults search for information?

Careful selection of questions to ask is very important in many situations. Examples include a child asking a question to learn the meaning of a novel word, a scientist choosing an experiment to differentiate between competing hypotheses, or a person’s visual system directing the eyes’ gaze to informative parts of a visual scene.

The psychology and mathematics of information search

Many studies investigating information search have used variants of the “20-questions” game. The task is to identify an unknown target item by asking as few yes–no (binary) questions as possible. Younger children tend to ask about specific objects (e.g., “Is it Paul?” or „Is it the number seven?“), whereas older children and adults tend to ask about properties that differentiate between subsets of multiple objects (e.g., “Is the person wearing a hat?” or „Is the number larger than 9?“). Mathematically, devising an efficient search strategy equates to finding the question tree (binary decision tree) that has the smallest expected total number of questions. However, due to the large number of possible trees, identifying the optimal strategy can be impossible. One way to approach the problem is to reduce the uncertainty about the target item in a stepwise manner by only considering the immediate next question. Uncertainty can be formalized using the concept of „entropy“ from information theory. The usefulness of a question is its information gain – the more a question reduces uncertainty, the better it is.

The split-half heuristic

Psychological research on information search in the 20-questions game has found that many people (children and adults alike) use a simple strategy: They ask about a feature whose distribution is closest to 50:50. For instance, if half of the remainin persons in the set have a hat, and other features are less equally distributed, this strategy would ask about hat. It can be proven mathematically that under many circumstances the split-half heuristic identifies the most informative question. This adds to a growing body of research showing that simple (or fast and frugal) heuristics can be competitive with or even outperform more complicated methods in real-world search, inference and decision tasks.

From games to education tools

There is now agreement in Germany that the concepts of “algorithm”, “information”, “bit”, and “code” are important for primary and secondary education. A key question is how to introduce children to these concepts in an intuitive and helpful manner. The 20-questions game and related games (e.g. Mastermind) can be used for playful activities which foster young children’s intuitions. They can also be related to teaching proportions, probability theory, and statistics.