|
|
| Psychological Record, The: Utilizing a computerized video poker simulation for the collection of dat |
The current paper presents data collected utilizing a computerized video poker simulation that was published in Behavior Research Methods, Instruments, and Computers by Dixon, MacLin, and Hayes (1999). To date no published studies have demonstrated the utility of the simulation or presented the possible data that may emerge from using the program. Eleven undergraduates played the simulation for 100 trials each for course extra-credit and the opportunity to win additional monetary compensation. Results suggest that there are regularities within and across participants on both temporal and subjective dependent variables during their play of the video poker game. These variables included response latencies, decision-making times, and subjective probability estimations. Implications of these data for the successful treatment of pathological gamblers and suggestions for further behavioral research on gambling are discussed.
Gambling behavior is on the rise across America. Twenty years ago only 2 states had legalized gambling, while 48 outlawed it. Currently the opposite is true. Forty-eight states now allow some form of legalized gambling with only Utah and Hawaii continuing a ban on it (Ghezzi, Lyons, & Dixon, 2000). With the rise in accessibility to gambling comes an increase in the number of potential problems associated with this risktaking behavior. Persons with problem gambling behavior are more prone to commit suicide and co-exhibit rates of substance abuse, income related crime, and other legal troubles in higher proportions than the rest of the population (DeCaria, Hollander, Grossman, Wong, Masovich, & Cherkasky, 1996; Politzer, Morrow, & Leavey, 1985).
There have been many hypotheses regarding why an individual would engage in gambling behavior when the chances of winning are against them. It has been argued across many disciplines that people gamble because they seek sensation (Anderson & Brown, 1984), possess certain biases or heuristics (Kahneman & Tversky, 1972), suffer from personality disorders (Kroeber, 1992), or are equipped with poor evaluation skills (Gilovich, 1983). Yet within the behavioral community, a comprehensive behavior-analytic approach is lacking. Skinner's contribution to a behavioral analysis of gambling was that the behavior was maintained by the variable-ratio schedule in effect (Skinner, 1953). Knapp (1997) furthered and clarified Skinner's position by stating that it was actually the random-ratio schedule in effect that functioned to maintain gambling behavior. The difference here being that each game, play, or trial's outcome is independent from the last. Rachlin (1990) added to the analysis by suggesting that gamblers tend to discount their string of losses, and more saliently remember their infrequent wins.
Recent research from a behavioral perspective has provided additional insight regarding the potential variables responsible for maintaining gambling behavior. For example, Lyons and Ghezzi (1995) illustrated the irrational relationship between lottery jackpot size (and the resulting probability of winning) and the quantity of lottery tickets purchased across Oregon and Arizona. They claimed that as jackpot size rose in a given state, and as a consequence the odds of winning that jackpot became lower, lottery ticket sales increased. In other words, more people purchased lottery tickets as their odds of winning the jackpot became worse. A recent study by Dixon, Hayes, and Aban (2000) demonstrated the role that verbal behavior in the form of rules may also have on the tendency to continue playing when the odds of winning became worse. In their study, many participants were exposed to a fair roulette game that became considerably worse over trials in terms of the probabilities of winning. Participants given inaccurate rules about roulette played for longer periods of time and took higher risks than participants given accurate rules about the game although the programmed contingencies were identical across both groups of participants. Together these studies may suggest that in addition to the role of programmed reinforcement contingencies, it is possible that rule-governed behavior contributes to the maintenance of gambling behavior.
Other than the many psychological explanations for persistent gambling behavior, there are also increasing numbers of physiological (e.g., Carlton & Goldstein, 1987) and genetic theories (e.g., Slutske, Eisen, True, Lyons, Goldberg, & Tsuang, 2000). These deviations towards a medical model for an observable behavior may eventually result in nonpsychological and/or drug-based treatments as the treatments of choice. Therefore, it may be a critical time for psychological researchers, and specifically behavior analytic researchers, to demonstrate the specific environmental contingency arrangements that may be controlling gambling behavior. Although it may be true that the random-ratio schedule maintains responding, to date we know nothing more about how the many parameters of that schedule influence behavior or the subtleties in responding that may occur within a gambling session.
The video poker software described by Dixon and colleagues (1999) was a mechanism designed to allow behavioral researchers to collect detailed data on gambling behavior. Like poker machines found in a casino, the simulation utilized a 52-card deck of 5-card draw poker. Participants could play the game identically as they would in the casino while earning class extra points or small monetary consequences for participation. Yet in contrast to an actual video poker game, the simulation allowed the experimenter to manipulate a number of game parameters (i.e., maximum bet/hand, estimations of winning, number of trials played, etc.) and collect response data on a trial-by-trial basis (response latency, engagement time, decision making time, etc.) (see Dixon et al., 1999 for a full description of the video poker program).
The software program was offered free of charge by the authors to any party interested in collecting such data. Yet, to date, no published data have resulted from the utilization of this program. Therefore the purpose of the present study was to utilize the software of Dixon et al. (1999) for the collection of behavioral data of undergraduate students wagering points exchangeable for class extra-credit and the opportunity to receive a cash prize. We subsequently assessed the resulting order in participants' response patterns from the various contingency arrangements operating during the course of video poker play.
Method
Participants
Eleven undergraduate students who had experience playing either video poker or traditional poker games participated in this study All participants were recruited from courses and were offered extra credit and the opportunity to win a $50 certificate to the campus bookstore for participating. Participants were told that the $50 certificate would be awarded to the player with the highest score of all players on the video poker game. All participants signed an informed consent form before participating.
Apparatus and Setting
The experimental apparatus consisted of an IBM compatible microcomputer equipped with a mouse and speakers. Programming of the video poker simulation was completed with Microsoft Visual Basic 5 for Windows. Upon initiation of the simulation, the program allows the experimenter to select among a variety of parameters for use during run time. The following were selected for the present experiment: Subjective probability = always on; Number of trials to play = 100; Maximum bet = 5 points; Show credits = yes; Sounds = on; Reinforcement magnitude = default settings. All sessions were conducted in a small quiet room approximately 10' by 11'. These parameters resulted in a fair game where all outcomes (i.e., wins or losses) were completely due to chance.
Procedure
The participant began the simulation by reading instructions and learning how to use the computer interface. Specifically they were told:
Before each trial a probability bar will appear. Use the bar to indicate how confident you are that your next game/trial will be a winning one. Selecting a'1' indicates that you guess the next hand you receive will be a losing hand for sure, while selecting a '10' indicates that your next hand received will be a winning hand for sure. Respond on the numbers between 1 and 10 to your varying degree of confidence about the outcome of your next trial. Try clicking on a few numbers below.
The next screen provided more instructions about the deck of cards and the procedure for each trial.
Continued from page 1.
During the game, once you select a probability number, the next hand will begin. You will need to click on the money icon located on the upper right corner of the screen for each coin credit you wish to bet. Once you have made your bet, deal cards by clicking on the 'deal' button. Hold the cards that you wish to keep by clicking on the 'hold' button. To cancel the cards held, click on the 'cancel' button. To draw new cards click on the 'draw' button.
Once the participant had completed reading the instructions, clicking on the start button began the game. Figure 1 provides a sequential display of participant behavior for a given trial, while Figure 2 displays the graphical interface. As can be seen in Figure 1, the first response was to click "continue" and initiate the trial, tn. Next, the participant chose a subjective probability value or the likelihood of winning on the upcoming trial. A response was then made for the amount of coins wagered. After the bet was made, the participant clicked on the "Deal" button and five cards appeared face up on the screen. The participant then chose which cards) were held, by clicking on the "Hold" button below that specific card. After the cards were chosen the participant clicked on the "Draw" button (even if all cards are held). The new cards appeared and the number of credits won or lost (positive or negative) appeared in the left hand corner of the screen. Finally, the participant clicked "continue" to proceed with the next trial. Participants were required to respond for 100 trials at which time the simulation program stopped.
Upon completion of the 100-trial experiment, participants were asked the following question:
Think back to your play during the game. You just played a total of 100 hands of video poker. Approximately, what was the total number of trials/hands that you won on?
Results
Figure 3 displays the mean response latency in seconds for each participant. These times were calculated by subtracting the mean continue time trial t^sub n-1^ from draw time trial t^sub n^ for each participant. All 11 participants demonstrated longer response latencies following winning trials, or reinforced trials, than following losing trials, or nonreinforced trials. Of those 11 participants, 9 had at least 2-s differences across trial types. A paired t test was conducted to support visual inspection of the data. A significant difference was observed between the two times [t(10) = -7.053, p
Figure 4 displays the mean decision time in seconds for each participant. Decision times were calculated by subtracting mean draw time trial t^sub n^ from deal time trial t^sub n^ for each participant. Of the 11 participants, 9 demonstrated shorter decision times preceding winning trials, or reinforced trials, than preceding losing trials, or nonreinforced trials. Of those 9 participants, 7 had at least .5-s differences across trial types. A paired t test was conducted to support visual inspection of the data. A significant difference was observed between the two times [t(10) = 3.018, p
Figure 5 displays the mean subjective probability after a winning trial for each participant. Mean subjective probabilities were calculated by computing the mean estimate separately for winning trials and for losing trials for each participant. This figure indicates that 9 out of 11 participants made higher estimates after a winning trial than after a losing trial. That is, participants reported that they were more likely to win after a winning trial and more likely to lose after a losing trial. A paired t test was conducted to support visual inspection of the data. A significant difference was observed between the two estimates [t(10) = 2.407, p
Figure 6 displays each participant's response to the postexperimental question asking them to estimate the number of winning trials they contacted. Each participant's actual number of winning trials is plotted for comparison. These data indicate that 9 of the 11 participants made trials estimations lower than they actually contacted, while only 2 made estimates higher than contacted. Deviations between their subjective estimates and the actual trial numbers were no less than 13 trials for any participant. A paired t test was conducted to support visual inspection of the data. A significant difference was observed between the two estimates [t(10) = -2.34, p
Discussion
Following a winning trial, and consequently the delivery of a reinforcer in the form of points, all 11 participants displayed longer response latencies when compared to following a losing trial. Similar findings have been obtained with participants engaged in slot machine play (Schreiber & Dixon, 2001), which suggests the present results extend this observation to video poker play. More important than the sheer difference across trial types was the current observation that participants' response latencies shortened as number of nonreinforced trials increased. These data might be conceptualized through a negative reinforcement and avoidance paradigm (Hineline, 1977). Here a losing streak in gambling increases in aversiveness as it becomes longer. Each subsequent losing trial may have additional aversive functions for the player, and as a result they initiate the onset of the next trial to escape the continued presentation of that aversive stimulus. Another possible conceptualization might be based on the notion of the establishing operation (Michael, 1993). Here, a participant who is experiencing increasing levels of deprivation (lack of a winning hand) responds at faster rates in attempts to obtain a potential reinforcer. Following reinforcer consumption (i.e., a win), the person has reduced the level of deprivation which results in a subsequently slower response rate on the next trial. It is often reported that many problem gamblers immediately seek to "win back" their recent losses by returning to the casino the next hour or day. The present findings and interpretations may suggest the foundations for a more objective analysis of this "win back" phenomena.
Unlike slots or roulette, video poker has a distinct "decision time" where the participant must choose to hold or discard zero through five of the five dealt cards initially comprising his/her hand. The present study illustrated temporal regularities in this decision time for most participants across winning and losing trial types. Specifically, participants tended to spend more time engaged in selecting which cards to hold and discard when the outcome of that trial was a loss compared to when the outcome of that trial was a win. Two conceptualizations of this finding are provided. The first conceptualization is again based on the potential aversive properties that an impending losing hand may have for a participant. In this case in the presence of an S-Delta for a win (i.e., the probably losing cards), the participant postpones the presentation of that aversive event by not responding on the draw button to complete the trial. Yet, the potentially reinforcing properties of the upcoming winning hand in the presence of an SD for a win (i.e. the probable winning cards) might result in the participant emitting shorter decision times to gain more rapid access to that reinforcer. The second conceptualization is based on problem solving (Donahoe & Palmer, 1994; Nakajima & Sato, 1993). Here it is quite possible that the participant is presented with a more complex problem or task on trials that eventually result in a loss, resulting in longer decision times. Here the participant is attempting to solve the problem of which nonwinning cards should they hold and discard. They may be attempting to calculate the odds of each type of winning hand given their available cards along with the odds of improving upon their hand with the next draw. Yet, given a relatively better hand of five initial cards, the problem becomes easier. Here the participant can immediately select to hold the winning cards and discard all others. Less decision making time is necessary.
Continued from page 2.
An unexpected finding from the current study was that the average subjective probabilities after reinforcing and nonreinforcing trials (wins and streaks of losses) were in direct contrast to nonbehavioral hypothesized notions of a "gambler's fallacy" (Yackulic & Kelly, 1984). The fallacy here is that gamblers tend to believe that following a winning trial it is more likely that a deviation in the opposite direction would occur on the next trial (i.e., a losing trial). The present data suggest that this fallacy may not hold unconditionally. Only 2 of the current 11 participants demonstrated mean subjective probability estimates consistent with the "gambler's fallacy." Of greater importance though is that all 11 participants in the current study varied their estimates across trial types. Objectively, each trial's outcome is independent of the last, the one before that, and the one before that. In other words, the video poker machine, or any gambling device for that matter, does not have a history. Each game operates on a random ratio or probability schedule of reinforcement. If the current participants were aware of this game characteristic all subjective probabilities should have been constant across trial types for each participant. Previous gambling research has demonstrated that more accurate estimates of probabilities can be taught to participants, yet without specific training these estimates will deviate substantially from objective probabilities (Dixon, 2000). Subjective probabilities are just that, subjective, and may need to be looked at in relation to each person's experience with gambling and how he or she individually assesses the probability of winning. Trends over time and across participants may appear, but initial analyses may need to be individual. Obviously, more research needs to be conducted in this area.
During postexperimental questioning, most participants (9 of 11) tended to underestimate their wins and overestimate their losses. One explanation for this finding is the participants may have followed a general rule that stated they would not win very often at gambling and chose values consistent with that rule while failing to reference their actual experience during the trials. Future research may wish to incorporate a direct analysis of participants' rule following repertoires before the onset and/or during such gambling tasks. Another explanation may be that losing trials were more salient to participants than winning trials and were therefore more easily recalled. This analysis also requires further investigations possibly integrating participant ratings of types of trial outcomes.
Awareness of the characteristics of gambling, such as these described in the present study, may provide an insight into possible intervention strategies for problem gamblers. If researchers conceptualize the longer response latencies and decision-making times as breaks in the temporal stream of play, perhaps clinical interventions could be targeted directly at this point. For example, during these pause times, problem gamblers could be instructed to assess how many coins they have won and how many coins they spent acquiring those winnings. With effective estimation tactics in place, problem gamblers may become more aware of their actual losses, make more rational estimates of winning and subsequently terminate play.
We feel it important to note that the simulation program of Dixon et al. (1999) collects additional data from what has been reported here in the present study. These additional forms of data can provide future researchers with additional avenues of investigation. For instance, the actual cards dealt each hand are written to the data file on each trial and stored in numeric form. A macro or syntax can be written to convert the cards from numeric to alpha numeric (e.g., 1113 = Jh, Jack of Hearts). Such recoding would allow a researcher to analyze card decision choice and also for additional analysis of risk within the game. For example, one might assess the frequency of a player's discarding of high probability winning cards with a low magnitude of payoff (i.e., a Queen and a Jack) for the opportunity to obtain new cards with a lower probability of winning, yet at a higher magnitude of payoff (i.e., a Royal Flush). Other deviations from optimal video poker play can also be analyzed through this type of recoding.
There are a number of additional research strategies that could be coupled with the poker program, so that together, they could provide further insights into gambling behavior. For example, participants could be asked to "think aloud" (Erricson & Simon, 1984) during their play of the game. The resulting data could then be examined for reasoning patterns, accurate and inaccurate self-rules, and/or vocal intensity changes across trials and strings of losses. Physiological data such as heart rate, temperature, or muscle tone might also be collected in attempts to assess the interaction between reinforcement and physiological arousal. Other supplementary strategies might include personality inventories, family histories, and assessments of previous risk-taking behaviors.
In summary, the observations from this study indicate that there are regular patterns of temporal responding that occur during the course of video poker play across most participants. Additionally there are considerable differences between participants' verbal estimations of winning and actual wins and losses. Together, these data suggest that both programmed contingencies and verbal descriptions of those contingencies interact when a human organism is engaged in a game of chance. It currently appears that this behavior is relatively predictable, and we hope, eventually controllable. As behavior analysts become more aware of the environmental arrangements that participate in sustaining gambling, they may be better prepared for designing effective clinical interventions based on their science.
References
ANDERSON, G., & BROWN, R. I. F (1984). Real and laboratory gambling, sensation-seeking and arousal. British Journal of Psychology, 75, 401-410.
CARLTON, P L., & GOLDSTEIN, L. (1987). Physiological determinants of pathological gambling. In T. Galski et al. (Eds), The handbook of pathological gambling. Springfield, IL: Charles C Thomas.
DECARIA, C. M., HOLLANDER, E., GROSSMAN, R., WONG, C. M., MASOVICH, S. A., & CHERKASKY, S. (1996). Diagnosis, neurobiology, and treatment of pathological gambling. Journal of Clinical Psychiatry, 57, 80-84.
DIXON, M. R. (2000). Manipulating the "Illusion of Control": Variations in risktaking as a function of perceived control over chance outcomes. The Psychological Record, 50, 705-720.
DIXON, M. R., HAYES, L. J., & ABAN, I. (2000). Examining the roles of rule following, reinforcement, and preexperimental histories on risk-taking behavior. The Psychological Record, 50, 687-704.
DIXON, M. R., MACLIN, 0. H., & HAYES, L. J. (1999). Towards a molecular analysis of video poker playing. Behavior Research Methods, Instruments, and Computers, 31(1), 185-187.
DONAHOE, J. W., & PALMER, D. C. (1994). Learning and complex behavior Boston, MA: Allyn and Bacon.
ERICSSON, K. A., & SIMON, H. A. (1984). Protocol analysis: Verbal reports as data. Cambridge, MA: MIT Press.
GHEZZI, P., LYONS, C., & DIXON, M. R. (2000). Gambling from a socioeconomic perspective. In W. K. Bickel & R. E. Vuchinich (Eds.), Reframing health behavior change with behavioral economics. New York: Erlbaum.
GILOVICH, T. (1983). Biased evaluation and persistence in gambling. Journal of Personality and Social Psychology, 44(6), 1110-1126.
HINELINE, P. N. (1977). Negative reinforcement and avoidance. In W. K. Honig & J. E. R. Staddon (Eds.), Handbook of operant behavior. New York: Prentice Hall. KAHNEMAN, D., & TVERSKY, A. (1972). Subjective probability: A judgement of representativeness. Cognitive Psychology, 3, 430-454.
KNAPP, T. J. (1997). Behaviorism and public policy: B. F. Skinner's views on gambling. Behavior and Social Issues, 7, 129-139.
KROEBER, H.- L. (1992). Roulette gamblers and gamblers at electronic game machines: Where are the differences? Journal of Gambling Studies, 8, 79-92. LYONS, C. A., & GHEZZI, P. M. (1995). Wagering on a large scale: Relationships
between public gambling and game manipulations in two state lotteries. Journal of Applied Behavior Analysis, 28, 127-141.
MICHAEL, J. (1993). Establishing operations. The Behavior Analyst, 16, 2-24. NAKAJIMA, S., & SATO, M. (1993). Removal of an obstacle: Problem-solving behavior in pigeons. Journal of the Experimental Analysis of Behavior, 59, 131-145.
POLITZER, R., MORROW, J. S., & LEAVEY, S. B. (1985). Report on the cost benefit/effectiveness of treatments at the John Hopkins Center for Pathological Gambling. Journal of Gambling Behavior, 2, 131-142.
RACHLIN, H. (1990). Why do people gamble and keep gambling despite heavy losses. Psychological Science, 1, 294-297.
Continued from page 3.
SCHREIBER, J., & DIXON, M. R. (2001). Temporal characteristics of behavior on random-ratio schedules observed during slot machine play. Psychological Reports, 89, 67-72.
SKINNER, B. F. (1953). Science and human behavior. New York: Macmillan. SLUTSKE, W. S., EISEN, S., TRUE, W. R., LYONS, M. J., GOLDBERG, J., &
TSUANG, M. (2000). Common genetic vulnerability for pathological gambling and alcohol dependence in men. Archives of General Psychiatry, 57(7), 666-673.
YACKULIC, R. A., & KELLY, I. W. (1984). The psychology of the "Gambler's Fallacy" in probabilistic reasoning. Psychology - a Quarterly Journal of Human Behavior, 21(3-4), 55-58.
MARK R. DIXON and JAMES B. SCHREIBER
Southern Illinois University
Address all correspondence to Mark R. Dixon, Behavior Analysis and Therapy Program, Rehabilitation Institute, Southern Illinois University. Carbondale, IL 62901. (E-mail: mdixon@siu.edu).
Copyright Psychological Record Fall 2002
Provided by ProQuest Information and Learning Company. All rights Reserved
|
|
