2. A Simple Decision-Making Task
We developed a simple, custom-built decision-making game called Buckets. It was designed to allow for the controlled collection of performance data, something that can be difficult in commercial games due to the complexity of interactions. Our Buckets game consists of a repeated task that was initially designed as a visual-only perceptual challenge for measuring how players employ strategy, balancing the risks and rewards associated with game mechanics (Williams, Nesbitt, Eidels and Elliott 2011).
Balancing risk and reward is an important consideration in the design of computer games and has even been likened to the thrill of gambling (Adams 2009). Of course, if players gamble on a strategy, they assume some odds, some amount of risk. In gameplay it is reasonable to expect that greater risks will be compensated by greater rewards. Adams (2009) not only states that each “risk must always be accompanied by a reward” but also describes this as a fundamental rule for designing computer games.
In the Buckets game, players must solve a perceptual challenge, deciding which of four rectangles (buckets) is filling up with dark blue dots (rain) the fastest. The game is comprised of repeated trials. With each trial, a new display of four buckets appears, and the player has one attempt to determine by a key press which of the four items is the target. A new trial with a fresh display appears after a response, irrespective of whether it was correct or not. The player’s overall goal in the game is to identify as many target buckets as possible within a fixed time period. Speed is important; however, the longer a player waits during each individual attempt (trial), the more likely it is that the attempt will result in a positive outcome. This is because as each trial progresses, more pixels accumulate in the target bucket, making it easier to discern the one correct bucket from the three incorrect ones. Thus, while a typical strategy might be to attempt faster responses, as this will reward the player with more time for additional attempts, an alternative strategy is to reduce the risk of each attempt by waiting longer to improve one’s chances of correctly identifying the target bucket.
Four buckets are displayed at the start of each trial (see Figure 4). Each bucket consists of 5,000 pixels (50 wide x 100 high), and each of the four buckets is 50% filled with 2,500 dark blue pixels and 2,500 white pixels. The actual position of the 50% dark pixels is chosen randomly at every update of the display. For each trial one of the four buckets is randomly chosen to serve as the target. Each decision must be made within the 8 seconds, total, of the trial. During this time the target bucket will gradually increase its number of blue pixels until it is 52.5% filled, while the other buckets are filled at a slightly slower speed. Determining the target bucket still remains a difficult perceptual challenge. This number (52.5%) was chosen by trialing the game and using the empirical data to set a difficulty level that gave players a 40-60% chance of success. If no response is given within 8 seconds, a timeout occurs, no score is given to the player, and a new trial begins.
Figure 4: Buckets Perceptual Challenge. The player responds by selecting one of the four keys {a, s, ;, ‘}
The frame rate of the game is configured to ensure that the display is updated at 10 frames per second. This means that about 14 extra dark blue pixels are added to the target bucket at each frame. The actual frame rate of the game is monitored to ensure it meets this required number of frames per second.
As described above, a player has only 8 seconds to respond; if they wait too long, they timeout, losing the opportunity to make a selection. After collecting initial empirical data for the game, we found an unexpected consequence of the design was that many players were experiencing these timeouts. This negated the benefit that was meant to accrue by waiting longer to make a decision. The risk of timing out thus had a disproportionate impact on the reward of waiting longer. To address this timeout problem, a simple sound alarm was designed. A beep was used to warn the player that they had 2 seconds left to respond. Unfortunately, when trialing this solution, the alarm proved distracting to some players, diverting them from their primary task and prompting them to make an immediate decision rather than allowing them to maintain focus on their primary perceptual challenge.
While such auditory alarms are commonly used as warnings, they are intended to divert user attention away from their current task (Patterson 1982; Walker and Kramer 2006). However, peripheral sounds have also been found useful for background monitoring of system states. We therefore implemented a simple background sound that increased in amplitude over the 8 seconds of the task decision. This sound can be described as an auditory icon, as the increasing sound acts in a way similar to the sound of an approaching car. The sound becomes louder (indicating more danger) towards a critical moment in time. There are some contraindications for using amplitude in this way (Brewster, Wright and Edwards 1993), so we also considered increasing the frequency of the sound. This, however, required us to resolve the complex relationship between pitch and amplitude (Walker and Kramer 2006), something that was difficult to resolve in the software platform we used. Initial pilot trials with this increasing amplitude sound indicated that a simple monitoring signal was suitable, and therefore we adopted this approach for further empirical testing.