Decision Maker, Part 1

Now that I’ve explained our lab’s method of game development and rapid prototyping, I’m going to briefly explain the rationale for some of the games in development and describe the progress we’ve made in the past two weeks. Keep in mind that we only have six weeks to design the games, collect data, and present our results at a local conference. It’s a sprint.

Decision Maker is designed to teach students about decision-making and to help them make better decisions under uncertain conditions. Decision-making was thoroughly studied by Danniel Kahneman and Amos Tversky, who are considered the fathers of behavioral economics. Kahneman received a Nobel Prize for their work in 2002 (after the passing of Tversky in 1996). Their model of decision-making was called Prospect Theory and, in the classic paradigm, subjects must choose between a sure bet (e.g., $100) and the prospect of winning a lottery (e.g., 50% chance of winning $214). The utility of each prospect is defined as the probability multiplied by the value. In our example, the prospect would be the wisest choice because the overall utility of the bet, $107, is higher than the utility of the sure bet, which is $100.  What Kahneman and Tversky discovered, however, was that people often behaved irrationally when presented with bets that had extreme probabilities or values. The astronomical prize money and the misperception of extremely low probabilities explain why people would pay $5 for a state Lotto ticket when the odds of winning are overwhelmingly against the player. While Prospect Theory has been around for years, there have been few efforts to shape behavior given this knowledge. Merely telling people how to make decisions is not enough to alter their behavior. Consequently, Decision Maker is designed to train students to improve their decision-making skills for extreme probabilities and values.

Designing games that are fun, educational, and able to collect data for scientific purposes is challenging. After the educational objectives of our games are established, we develop experimental protocols that will allow us to assess the behavior we wish to shape. Then, we develop game mechanics that compliment those experimental protocols. Our lab uses Tracy’s Fullerton’s Game Design Workshop and Jessie Schell’s Game Design: A Book of Lenses to ensure we address the most critical elements of game design. While it’s not necessary to adhere to this particular order, I recommend starting with a good scientific experiment. However, it is sometimes useful to develop the experiment and game mechanics in separate “sandboxes.” I often ask students to simultaneously design an experiment, a game mechanic, and a fun toy/interface to play with. Combining the results of independent endeavors often produces interesting and unexpected surprises.

Fortunately for Decision Maker, there are several experimental paradigms from behavioral economics to consider. We adopted a method of adjustment paradigm where players indicate the sure bet they would accept in lieu of a given prospect. Players will be presented with positive and negative prospects, and prospects will vary widely in probability and value. The set of probabilities and values were randomly generated to make the mental computation of utility difficult (e.g., 0.2% chance of winning $10,147). Using the staircase method typically employed in psychophysical experiments, the difficulty of the decisions will increase when players make more correct decisions, and the difficulty will decrease if they make errors. The staircase allows us to find the threshold where decisions become less reliable, and it keeps players in a state of flow (where the task is optimally challenging without being too frustrating or too boring). While multiple ascending and descending staircases can be interleaved, we decided to go with a simple descending staircase with consistent step sizes between trials. Our prediction is that subjects who play our game will preform better on a post-test of decision-making relative to control subjects who received equivalent practice in decision making without using games.

After the experiment was designed, we developed the game mechanics. Some of these mechanics are more clearly defined in Tracy’s book, but I’ll briefly define them here. Objectives describe what the player is trying to achieve in the long run. They can describe intermediate goals or the ultimate win/fail states of the game. Resources are items in the game that you are either acquire or get rid of to achieve your objective (e.g., the pieces in Chess or the money in Monopoly). Feedback mechanisms are implemented in games to inform the player about their performance (e.g., point totals or badges). Reward/Punishment Contingencies describe how and at what rate the game will react to a player’s decisions. Different contingencies can have dramatically different effects on behavior. For example, more work is typically elicited from pigeons if rewards are intermittent rather than consistent. With effective contingencies and feedback mechanisms, behavior can be shaped quickly to help the player achieve their objective. Of course, games don’t always have to be friendly. Unreliable feedback mechanisms can be used to achieve a different effect. Flow has been defined in detail elsewhere. Students should focus intensely on how to use all the other mechanics and standard psychophysical procedures to elicit a state of flow in the players. The staircase method should be the starting point when considering flow. Finally, boundaries are rules that prevent players from acting in a particular way. While limiting player behavior might appear to be a fun killer, boundaries often have the opposite effect. For example, soccer is really only fun because players are not allowed to use their hands. This game mechanic is best employed when you are not making progress on a design. If a student is functionally fixed on a particular design that is not working, introduce a boundary to change the designer’s frame of reference.

The core game mechanic for Decision Maker rests on how prospects are presented to the player and how the player evaluates those prospects. Prospects will be presented on playing cards along with scenarios related to student life. Players must write down the sure bet they would accept in lieu of the prospect. The correct utility of the prospect will appear on the back of the card. Players are rewarded for correct decisions by receiving a card and they are punished by not receiving a card. The objective of the game is to accrue more cards than your opponent before the deck of cards is exhausted. After a choice is made, the player spins the dial of a spinner and watches the gamble play out in real time. It’s important to note that, just like the state Lotto, players can win on rare occasions even if the choice to gamble is incorrect. Immediate feedback is implicit when the player flips the card to see the correct answer, but feedback is also available by comparing how many cards each player has collected relative to the opponents. We didn’t really feel the need to impose boundaries in this game because the behavior is fairly controlled and we didn’t want to further limit our players. Because we are developing this game as a board game, implementing a psychophysical staircase procedure was a little trickier. We decided to introduce levels into the game. Each level will have it’s own spinner and deck of cards. During early levels, the player will evaluate relatively easy prospects (i.e., within the range of reliable decision making). If the player successfully answers a certain number of questions, they are advanced to the next level where more difficult decisions have to be made. If a player doesn’t accrue a significant number of cards at the end of a level, they must go back and replay that level. Thus, student must improve on their evaluation of difficult prospects in order to win the game. Spinners for the early levels will be marked to indicate probabilities from 1 to 100%. Spins for the advanced levels will represent extreme probabilities by requiring the player to make several spins in a row within a target zone (e.g., the bet for a 0.25% prospect pays off if the player gets the needle to land between 0 and 5 twice in a row). The physical task of spinning the spinner several times for low probabilities will hopefully reinforce the notion that it’s unwise to bet on rare outcomes! Players will keep track of their own score. Players must maximize their winnings and finish with the most number of cards to win the game. If a player finishes the game without having the largest total on their scorecard, all players purge their cards and the final level will be repeated until there is a winner.

To make the game more fun, we plan to add a number of physical tasks that also must be completed before advancing to the next level (e.g., stack all the blocks that come with the game Jenga). Players will have a limited time to complete mini-games, physical feats, or puzzles to pass to the next level. The addition of these puzzles will add variety to the game and allow players to take a break from performing mental calculations. Additionally, some game cards will introduce “windfalls” or “calamities” into the mix. Windfalls might spontaneously grant the player an extra turn, extra cards, or allow them to circumvent the physical task. Calamities might require the player to loose a turn, give up cards, or move back a level. To maintain balance between competitors, players with fewer cards are more susceptible to windfalls and players with more cards are more susceptible to calamities.

Our plan is to have a working prototype of this game in the next week and play test all our games so that we can collect data the following week. Subsequent posts will describe our other games and the progress made in this game.

Leave a Reply