Training gamblers to re-think their gambling choices: How contextual analytical thinking may be useful in promoting safer gambling.

BACKGROUND AND AIMS: Harmful gambling has been associated with the endorsement of fallacious cognitions that promote excessive consumption. These types of beliefs stem from intuitively derived assumptions about gambling that are fostered by fast-thinking and a lack of objective, critical thought. The current paper details an experiment designed to test whether a four-week online intervention to strengthen contextual analytical thinking in gamblers is effective in changing gamblers cognitions and encouraging safer gambling consumption.
METHODS: Ninety-four regular gamblers who reported experiencing gambling-related harm were randomly allocated to either an experimental (n = 46) or control condition (n = 48), including 45 males, ranging from 19 to 65 years of age (M = 36.61; SD = 9.76). Following baseline measurement of gambling beliefs and prior week gambling consumption, participants in the experimental condition were required to complete an adaption of the Gamblers Fallacy Questionnaire designed to promote analytical thinking by educating participants on common judgement errors specific to gambling once a week for four weeks. Post-intervention measures of beliefs and gambling consumption were captured in week five.
RESULTS: The experimental condition reported significantly fewer erroneous cognitions, greater endorsement of protective cognitions, and reduced time spent gambling post-intervention compared to baseline. The control group also reported a reduction in cognitions relating to predicting and controlling gambling outcomes.
CONCLUSION: Cognitive interventions that encourage gamblers to challenge gambling beliefs by reflecting on gambling involvement and promoting critical thinking may be an effective tool for reducing the time people invest in gambling activities.

Introduction

Gamblers endorse a variety of cognitive distortions that promote risky or excessive gambling ( Blaszczynski & Nower, 2002; Goodie & Fortune, 2013; Miller & Currie, 2008 ). The types of beliefs people have and the strength with which they endorse them are related to a person's preferred style of thinking ( Aarnio & Lindeman, 2005; Bloom & Weisberg, 2007; Gervais, 2015; Pennycook, Cheyne, Seli, Koehler, & Fugelsang, 2012; Swami, Voracek, Stieger, Tran, & Furnham, 2014; Willard & Norenzayan, 2013 ). Gamblers are more likely to demonstrate a preference for intuitively driven decision-making over more reflective or deliberative processing methods ( Armstrong, Rockloff, Browne, & Blaszczynski, 2019a; Cosenza, Ciccarelli, & Nigro, 2019; Emond & Marmurek, 2010 ), and as a result, tend to make decisions using mental processing shortcuts that appear to be designed to generate responses quickly and effortlessly ( Epstein, 2008; Hammond, 1996; Kahneman, 2003; Sadler-Smith & Shefy, 2004; Sadler-Smith, Zhang, & Sternberg, 2009 ). When applied to gambling, fast-intuitive thinking and a lack of critical reflection can impair judgement by strengthening the endorsement of erroneous gambling cognitions, that in turn contribute to poor gambling decisions ( Armstrong, Rockloff, & Browne, 2020; Armstrong, Rockloff, Browne, & Blaszczynski, 2020; Cosenza, Ciccarelli, & Nigro, 2019; Emond & Marmurek, 2010; Leonard, 2018 ).

A number of studies exploring non-evidence based beliefs have found that priming analytical thinking can weaken the strength with which beliefs are endorsed ( Gervais & Norenzayan, 2012; Paxton, Ungar, & Greene, 2012; Swami, Pietschnig, Stieger, & Voracek, 2011; Swami, Voracek, Stieger, Tran, & Furnham, 2014; Uhlmann, Poehlman, Tannenbaum, & Bargh, 2011 ). However, when applied to gambling, similar primes have failed to reduce cognitive distortions amongst gamblers ( Armstrong, Rockloff, Browne, & Blaszczynski, 2019, b ). In fact, Armstrong et al. (2019b) found that generalised analytic priming that involved a scrambled sentence task using keywords that facilitate analytical thinking (e.g., analyse, reason, ponder, think, rational) was counterproductive and actually increased positive gambling expectancies compared to a control condition. In some cases, gamblers may use analytical thinking to generate false narratives that justify gambling decisions, rather than seeking evidence to invalidate them ( Armstrong et al., 2019b; Boudry & Braeckman, 2012; Ellerby & Tunney, 2017; Kahan, 2018 ).

Designing an effective intervention that neutralises the formation and reinforcement of biased gambling cognitions, and that can be implemented beyond a clinical setting, is likely to be challenging. Given the tendency for gamblers to use critical or elaborative thinking in a way that rationalises or justifies poor gambling choices ( Armstrong et al., 2019b; Boudry & Braeckman, 2012; Ellerby & Tunney, 2017; Kahan, 2018 ), interventions designed to encourage analytical thinking must be designed with care to avoid being counterproductive. The task must elicit strong cues to the underlying rules associated with the problem, and increase vigilance in monitoring or deciphering the correct response ( Kahneman, 2003 ). For example, simply altering the framing of a task (e.g., a statistical test rather than a lottery) is enough to cue underlying rules at play and generate greater vigilance, resulting in stronger analytical responses ( Kogler & Kühberger, 2007 ). Interventions that are contextually relevant and serve to illustrate how gamblers are “tricked” into making biased decisions may be more effective in altering people's gambling cognitions, and in turn, their gambling behaviour ( Houdé et al., 2000 ).

Previous research exploring the effectiveness of long-term training in gambling odds and probability was ineffective in generating behavioural changes ( Williams & Connolly, 2006 ). This may be in part due to the nature of the training and sample characteristics. The training involved using gambling examples and themes in the provision of an education program teaching probability and odds. The aim was for students to learn the statistical nature of gambling rather than apply this learning to a gambling context. The sample consisted of university students, and while 71% reported to have gambled in the previous 6 months, majority of these had spent little time and money doing so and thus had no real motivation or reason to change their gambling behaviour.

The current paper describes the results of an experimental study exploring the effectiveness of a four-week online intervention designed to encourage gamblers to apply rational thought and statistical knowledge to overcome common gambling biases. The purpose was to determine whether training gamblers to think more analytically by solving a series of questions relating to common gambling biases developed based on the Gamblers Fallacy Questionnaire ( Leonard, Williams, & Vokey, 2015 ) and providing performance-based feedback would be effective in reducing gambling related cognitive distortions, and subsequently, decrease real-world gambling consumption amongst regular gamblers who experience gambling related harm. It was hypothesised that gamblers in the experimental condition would report fewer erroneous and more protective gambling beliefs and have lower real-world gambling consumption at the post-intervention follow-up compared to baseline and the control condition.

Method

Participants

A brief survey screen was released via online resource provider, Mechanical Turk, to respondents with hit approval rates of >96% and a minimum of 500 completed hits. To be included in the study, participants were required to provide informed consent, be over 18 years and reside in the United States, have gambled at least weekly on any form of gambling in the past 6 months (excluding lotteries, instant scratch tickets, or raffles), and score a 1+ on The Short Gambling Harms Screen ( Browne, Goodwin, & Rockloff, 2018 ). Participants who met the inclusion criteria were then randomly allocated to either a control or experimental condition. Of the 445 workers who completed the screening survey, 180 gamblers met the inclusion criteria, with 150 continuing to complete the baseline survey. Thirty-five cases were removed for having multiple entries or incomplete responses to the baseline survey, and 21 were removed due to attrition (failing to complete at least two of the weekly surveys or the post-intervention survey).

The final sample consisted of 94 participants (46 experimental; 48 controls), 45 males and 48 females (1 case identified as “other”), ranging from 19 to 65 years of age ( M = 36.61; SD = 9.76 ). The majority of participants were employed full time or self-employed (78.2%, 11%; respectively), most had either a bachelors degree (34.7%) or completed some college but had no degree (24.8%); and had a median personal annual income of between $600 and $799 USD (21%).

Based on responses to the Problem Gambling Severity Index (PGSI; Ferris & Wynne, 2001 ), 46% met the criteria for problem gambling, 29% for moderate-risk gambling and 9% and 10% were classified as low-risk or non-problem gamblers, respectively. Seventy-three participants (77.6%) exceeded the threshold for safe gambling consumption as measured by the Consumption Screen of Problem Gambling (score of 4+; Rockloff, 2012 ); and reported experiencing on average 6.43 gambling harms out of the 10 captured by the SGHS ( SD = 2.69 ; Browne et al., 2018 ). The most popular form of gambling was electronic gaming machines (EGMs; 32.9%), followed by sports betting (21.2%), blackjack (17.1%) and poker (16%). EGMs were also the mode of gambling on which players spent the most time (31%), and money (30.9%).

Measures

Gambling beliefs

The Gambling Related Cognition Scale (GRCS; Raylu & Oei, 2004 ) and the Protective Gambling Beliefs Scale (PGBS; Armstrong et al., 2019a ) were used to measure participants cognitions about gambling. The GRCS requires participants to rate their level of agreement to 23 items on a 7-point Likert scale and provides five subscale scores that capture erroneous beliefs or cognitions: an illusion of control (IC), interpretive bias (IB), predictive control (PC), gambling-related expectancies (GE), and perceived inability to stop gambling (IS). Cronbach alpha for the 23 items (or total GRCS score) was 0.93.

The PGBS is a 10 item scale measuring participants level of agreement on a 4-point scale to statements concerning protective cognitions about gambling, including: gambling expectancies, the role of personal skill, and the nature of chance and probability. Cronbach alpha for the 10 PGBS items was 0.87.

Prior week gambling

Questions regarding prior week gambling asked participants to report the amount of time (hours and minutes) and money (USD) they spent gambling in the week prior to the survey and during a typical gambling session that same week.

Procedure

The study consisted of six waves of data collection, including baseline (week 0), 4× weekly surveys (week 1–4), and a post-intervention phase (week 5). The baseline survey was designed to capture pre-intervention measures of gambling beliefs and gambling participation, as well as broader gambling involvement (problem gambling severity, gambling consumption, and gambling preferences) and demographics (age and gender). Following survey questions, the participants were presented with the first intervention task. The task varied depending on condition allocation during the screening phase of the study.

The experimental condition involved completion of an analytical training task designed to educate participants on common judgement errors specific to gambling. The intervention task was an extended form of the Gambler's Fallacy Questionnaire ( Leonard et al., 2015 ), which traditionally consists of ten multiple choice questions designed to tap into common fallacies associated with gambling. An additional 40 items were developed that challenged people's knowledge of these common gambling fallacies based on the original GFQ items (see Appendix A ). Each intervention task consisted of ten items, requiring participants to select the correct response from several possibilities. Immediately following their response, participants were provided performance-based feedback informing them of whether they were correct and providing a detailed explanation of the reason(s) underlying the correct response. Once all questions had been attempted, they were given the opportunity to revisit the questions they answered incorrectly to provide a revised answer based on the feedback they had received.

The control group also received a set of ten questions; however, their questions assessed knowledge of general gambling trivia (see Appendix B ). Determining a correct response on the general gambling trivia questions only requires memory retrieval of factual knowledge, and thus presumably does not activate the same cognitive processes as tasks that require insight and problem solving skills ( Cabeza, Dolcos, Graham, & Nyberg, 2002; Metcalfe, 1986 ). Participants in the control condition were not provided any feedback (performance based or otherwise) on the general trivia questions/answers, and thus, the task should not have elicited analytically driven cognitive processing.

The weekly surveys were administered one week following baseline (week 1–4). The weekly surveys measured participants prior week gambling involvement and provided them with the relevant task depending on condition allocation (i.e., either the extended GFQ, or Gambling-trivia). In week five, participants received the post intervention survey which re-assessed participants gambling beliefs and prior week gambling, as well as some sociodemographic characteristics. Participants were encouraged to complete all surveys within 48 hours of being made available to minimise potential overlap in reports of prior week gambling. Those who had not completed the survey within the first 24 hours (approximately) were prompted with a reminder message. Participants received monetary compensation based on the length of each survey, with longer surveys (baseline and week five) offering greater incentives. Data collection commenced August 2019 and finished in September 2019.

Statistical analysis

The experiment was designed to test the impact of an analytic training task on gambling beliefs and gambling intensity. It was hypothesised that those who received the analytical training task would report fewer erroneous and more protective cognitions, and reduced gambling consumption (measured by a decrease in time and money spent gambling). Based on these predictions, the purpose of the analyses were twofold: a) to determine the extent to which the experimental manipulation changed the outcome variables from baseline (week 0) to post-intervention (week 5); and b) to compare the control and experimental conditions on the extent of change experienced during the experimental period of the study. As groups did not differ significantly in gambling beliefs or intensity at baseline, to test the latter, a difference variable was calculated for each outcome measure. This was calculated by subtracting week 5 scores from those obtained at baseline. Supplementary analyses explored changes in the 10 difference variables by PGSI category (problem versus moderate/low risk), but found no significant variation in change by group, P > 0.05, ns.

Ethics

The study procedures were carried out in accordance with the Declaration of Helsinki and granted formal ethics approval by Central Queensland University's Human Research Ethics Committee. All subjects were informed about the study and all provided informed consent.

Results

Gambling beliefs

A series of paired sample t-tests were conducted for each condition to compare baseline and week five measures of gambling beliefs. It was expected that the experimental condition would report fewer erroneous beliefs and more protective gambling beliefs at week five compared to baseline. As gambling beliefs are theoretically (and statistically) inter-related, a Bonferroni correction was applied to adjust for multiple related comparisons by dividing the P value of 0.05 by the number of comparisons being made (.05/6). The adjusted critical P value applied was 0.008.

As expected, the experimental condition had significant lower scores for: illusion of control, t (46) = 4.17, P < 0.001; interpretive bias, t (46) = 4.44, P < 0.001; predictive control, t (46) = 4.90, P < 0.001; inability to stop, t (46) = 2.88, P = 0.003; and protective gambling beliefs, t (46) = −2.39, P < 0.001. However, compared to baseline, the control condition also scored significantly lower on measures of illusion of control, t (48) = 3.54, P < 0.001; and predictive control, t (48) = 2.97, P = 0.002, at the week 5 follow-up. Table 1 presents the descriptive statistics and results of the paired sample t-tests for measures of gambling beliefs for each condition.

Table 1.

Descriptive statistics and Test of Equality for baseline and week five measures of gambling beliefs by experimental condition

		Baseline		Week 5		Test of equality
		M	SE	M	SE	t	M diff	SE diff	P
Control	IC	15.50	1.06	12.67	1.06	3.54	2.83	0.80	<0.001*
	IB	20.37	0.87	18.67	0.83	2.34	1.70	0.72	0.011
	PC	27.18	1.01	24.92	1.21	2.97	2.27	0.76	0.002*
	GE	20.25	0.72	18.97	0.83	1.80	1.27	0.70	0.039
	IS	16.92	1.20	15.25	1.25	1.96	1.67	0.86	0.028
	PGBS	27.90	0.90	29.20	0.87	−1.53	−1.31	0.86	0.066
Experimental	IC	14.00	0.99	11.30	1.01	4.17	2.69	0.64	<0.001**
	IB	19.58	0.64	16.53	0.74	4.44	3.06	0.69	<0.001**
	PC	24.89	1.27	20.47	1.12	4.90	4.41	0.90	<0.001**
	GE	19.77	0.65	19.06	0.59	1.34	0.69	0.51	0.092
	IS	16.94	1.23	14.56	1.08	2.88	2.37	0.83	0.003*
	PGBS	28.79	0.80	31.18	0.74	−4.02	−2.39	0.59	<0.001**

Note: Control n = 48, df = 47; Experimental n = 46, df = 45. * P < 0.008, ** P < 0.001 (1-tailed).

A Multivariate Analysis of Variance (MANOVA) was conducted to determine whether differences exist between the control and experimental condition in belief change from baseline to week five comparing difference variables. We expected that the experimental condition would report greater changes in gambling beliefs compared to the control condition from baseline to week five. Results showed there were no significant differences between experimental conditions for any changes in beliefs from baseline to week 5, except for predictive control. The experimental condition experienced a significantly greater change in scores for predictive control compared to the control condition, F (1.92) = 3.30, P = 0.036. Table 2 presents the MANOVA and descriptive statistics for difference variables for each condition.

Table 2.

MANOVA results comparing belief change by condition

	Multivariate		Univariate
	V	F		IC	IB	PC	GE	IS	PGBS
Condition	0.091	1.32		0.02	1.82	3.30	0.43	0.35	1.05
			P	0.447	0.090	0.037*	0.258	0.277	0.154
Control			M	2.83 (0.80)	1.70 (0.72)	2.28 (0.76)	1.28 (0.70)	1.67 (0.86)	−1.31 (0.86)
			(SE)
Experimental			M	2.70 (0.64)	3.06 (0.69)	4.41 (0.90)	0.70	2.37 (0.83)	−2.39 (0.59)
			(SE)				(0.51

Note: Control n = 48; Experimental n = 46; Multivariate sourced from Roys largest root, df 6, 87; Univariate df 1.92. * P < 0.05 (1-tailed).

Gambling intensity

Listwise deletion was used for values outside three times the standard deviation that were found to be incompatible when compared to other responses. Gambling intensity measures were highly positively skewed and could not be adequately corrected through transformation. Each outcome: minutes per week gambling, minutes gambling in a typical session, dollars spent per week gambling and dollars spent gambling in a typical session, were therefore analysed using nonparametric tests.

It was expected that the experimental condition would report lower gambling intensity at week five compared to baseline, and that changes in gambling intensity would be significantly greater in the experimental condition compared to controls. To determine whether gambling intensity changed from baseline to week five, a series of Related Samples Wilcoxon Signed Rank Tests were conducted for each condition. Mann-Whitney U Tests were used to test whether the degree of change between baseline and week five significantly differed between conditions for each gambling intensity outcome. As gambling intensity measures are likely to be related, a Bonferroni correction was applied (0.05/4) resulting in an adjusted critical P value of 0.012. Table 3 presents the descriptive statistics for gambling intensity measures.

Table 3.

Descriptive statistics for baseline and week five measures of gambling intensity by experimental condition

		Baseline				Week 5				Difference
		M	SE	Mdn	Range	M	SE	Mdn	Range	M	SE
Control	MPWG	223.25	22.10	215.00	600	190.22	23.85	165	650	14.96	29.04
	MGTS	93.20	8.36	90.00	250	91.56	10.01	90.00	245	−1.40	11.41
	DSPW	176.30	40.32	100.00	1,400	206.98	37.87	92.50	1,000	4.25	54.28
	DSTS	85.95	19.70	47.50	700	95.93	18.71	60	600	−8.35	26.35
Experimental	MPWG	230.83	32.44	180	750	163.21	23.57	130	720	66.22	24.09
	MGTS	109.33	12.83	97.50	360	76.30	11.48	60.00	300	33.03	15.59
	DSPW	159.45	29.50	80.00	700	166.50	39.79	50	1,000	−10.31	31.79
	DSTS	94.43	18.08	50	500	72.86	22.36	32.50	900	20.78	16.68

Note: Control n = 44, Experimental n = 42; Minutes per week gambling (MPWG); Minutes spent gambling in a typical session (MGTS); Dollars spent per week gambling (DSPW); Dollars spent gambling in a typical session (DSTS).

Minutes per week gambling

In the control condition, there were no significant differences between baseline and week 5 for minutes per week gambling, Z = −1.37, P = 0.085 (1-tailed). As expected, there was a significant difference in minutes per week gambling from baseline to week 5 for the experimental condition, Z = −2.47, P = 0.006 (1-tailed). Figure 1 below illustrates the frequency distribution of minutes per week gambling for the experimental condition at baseline and week five.

Fig. 1.

Minutes per week gambling frequency distribution for baseline and week five, experimental condition ( n = 45)

A Mann-Whitney U Test was used to determine whether there were differences between the experimental and control condition for changes in minutes per week gambling over the experimental period. However, the result was non-significant, Z = 0.28, P = 0.387 (1-tailed).

Minutes gambling in a typical session

There were no significant differences between baseline and week 5 scores for minutes spent gambling in a typical gambling session for the control condition, Z = 0.02, P = 0.492 (1-tailed). However, as expected, the experimental condition spent significantly less time gambling at week 5 compared to baseline during a typical gambling session, Z = −2.57, P = 0.005 (1-tailed). Figure 2 below illustrates the frequency distribution of minutes spent gambling in a typical session for the experimental condition at baseline and week 5.

Fig. 2.

Minutes spent gambling on a typical session frequency distribution for baseline and week five, experimental condition ( n = 42)

Results of a Mann-Whitney U Test showed there were no significant difference between the experimental and control condition for changes in minutes spent gambling in a typical session across the experimental period, Z = 1.68, P = 0.046 (1-tailed).

Dollars spent per week gambling

There were no significant differences in the total amount spent per week gambling between baseline and week 5 for the control condition, Z = 0.81, P = 0.210 (1-tailed), or the experimental condition, Z = 0.19, P = 0.422 (1-tailed).

A Mann-Whitney U Test demonstrated there was no significant difference between the experimental and control condition for changes in the total amount spent gambling per week across the experimental period, Z = 0.89, P = 0.187 (1-tailed).

Dollars spent gambling in a typical session

There were no significant differences in the amount spent on a typical gambling session between baseline and week 5 for the control condition, Z = 0.51, P = 0.306 (1-tailed), or the experimental condition, Z = −1.57, P = 0.058 (1-tailed).

Results of a Mann-Whitney U Test showed there were no significant difference between the experimental and control conditions in changes to the amount spent during a typical gambling session, Z = 1.56, P = 0.059 (1-tailed).

Discussion

The purpose of this study was to determine whether an online intervention designed to encourage analytical thinking and expose common gambling fallacies would be effective in generating more rational cognitions, reducing gambling intensity. While the intervention was successful in reducing erroneous beliefs across 4 of the 5 GRCS scales (illusion of control, predictive control, interpretive bias, inability to stop) and improving protective cognitions post-intervention, both conditions demonstrated fewer beliefs relating to predictive and illusionary control across the experimental phase. This change was significantly greater for the experimental group. These results suggest that the intervention: 1) made for stronger changes to beliefs relating to predictive control compared to the control group; 2) reduced endorsement of other erroneous cognitions (e.g., inability to stop gambling and interpretation of gambling outcomes; and 3) promoted safer cognitions about gambling.

Despite non-significant results for monetary expenditure, the intervention was effective in reducing the amount of time people spent gambling compared to baseline, however, changes in time spent gambling were not significantly different when comparing controls to the experimental condition. Prolonged training that challenges gambling fallacies may cause people to question their gambling choices, making gambling less enjoyable ( Lin, Hung, & Li, 2012; Wohl, Young, & Hart, 2007 ) and encouraging people to quit sooner. It has been suggested that reduced enjoyment should equate to greater risk aversion (e.g., smaller bets) ( Wohl et al., 2007 ), and since games of chance tend to have a negative expected value in the long term ( Walker, Litvin, Sobel, & St-Pierre, 2015 ), that a reduction of time spent playing would naturally equate to a reduction in gambling losses and reduced gambling expenditure. However, this was not the case in the current study. It may be that a reduction in the level of enjoyment as a result of more rational approaches to gambling may encourage people to gamble more money in an attempt to make gambling more exciting.

More likely, however, is that the lack of effect is due to how gambling expenditure was quantified. Participants were asked to report “In the last week, approximately how much money have you spent gambling in total” . Gamblers tend to have varying definitions or methods for calculating gambling expenses, with some including overall turnover and others considering only net expenditure ( Blaszczynski, Dumlao, & Lange, 1997; Wood & Williams, 2007 ). Ambiguous questionnaire items can therefore generate substantially different responses ( Blaszczynski et al., 1997; Wood & Williams, 2007 ). Further, those with gambling problems tend to lose track of money spent and have difficulty determining whether they are financially ahead during a gambling session ( Nower & Blaszczynski, 2010 ). Retrospective accounts of gambling expenditure are therefore likely to be unreliable for providing accurate information on the net value of money invested into gambling activities from personal bank rolls; a significant limitation of the current study. Nevertheless, it is important to note that lack of significant findings on expenditure do not imply that there was no change but may instead only reflect an inability to detect the change with self-report data. In future research, more objective measures, such loyalty card data, might provide an ability to detect a reduction in gambling expenditure. Logically, when intensity of play is constant, a reduction in time spent gambling should translate into a reduction in expenditure.

As well as the issues surrounding self-reported gambling expenditure measures, there are several other design elements that should be recognised as limitations of this study and considered when interpreting the findings. For instance, having participants monitor their gambling consumption may have prompted changes in perceived control over gambling outcomes, as demonstrated by results indicating changes to illusion of control and predictive control in both conditions. Further, the GRCS and PGBS have not been assessed for re-test reliability and the GRCS is not an exhaustive measure of gambling fallacies and biases. There may be other measures of cognitive distortions that better capture gambling cognitions that may be more or less affected by an analytic intervention. Findings of this study demonstrate the potential for cognitive interventions designed to challenge fallacious thinking by encouraging more analytic thought to be useful in attempts to reduce the time people spend gambling. However, given the intervention failed to change gambling expenditure, it would be premature to argue that the results would support such an intervention being adequate in generating and sustaining long term behavioural changes that reduces gambling related harm by itself.

In order for people to reduce gambling consumption and thus minimise their risk of experiencing gambling related harm, it is likely that cognitive interventions that challenge biased decision making would benefit from the addition of other strategies, such as behavioural feedback, that provides gamblers with realistic accounts of their actual gambling expenditure; allowing them to recognise behavioural patterns and moderate gameplay ( Wohl, Davis, & Hollingshead, 2017; Wood & Wohl, 2015 ). Recovering gamblers report many different techniques or strategies as helpful for minimising or abstaining from gambling ( Hodgins & el-Guebaly, 2000; Rodda et al., 2018 ). However, most studies that explore harm reduction techniques endorsed by gamblers tend to focus on behavioural strategies used by the gamblers, often overlooking the cognitive underpinnings that promote behavioural responses ( Hing, Nuske, & Gainsbury, 2012; Rodda et al., 2018 ). Interventions, such as the one used here, to strengthen decision-making skills that are based on reflection, critical thought and reality checking, are likely to provide greater control over gambling decisions, only increasing the effectiveness of behavioural strategies for reducing gambling consumption.

While cognitive techniques have been reported to be helpful in strengthening gambling abstinence ( Hodgins & el-Guebaly, 2000 ) and are incorporated into many problem gambling treatment therapies ( Fortune & Goodie, 2012; Gooding & Tarrier, 2009 ), they often require gamblers to access help services in order for them to be successful. Wohl et al. (2007) suggest that some biases, such as dispositional luck, are associated with negative attitudes towards treatment seeking, and thus, failing to address gambling related cognitive distortions is likely to be a barrier to help seeking. Further, many responsible gambling initiatives operate under the assumption that being an informed consumer translates to better decision making. However, research suggests that such approaches are largely ineffective at generating behavioural changes ( Cloutier, Ladouceur, & Sévigny, 2006; Miyazaki, Brumbaugh, & Sprott, 2001; Monaghan & Blaszczynski, 2005, 2007; Monaghan, Blaszczynski, & Nower, 2009; Williams & Connolly, 2006; Wynne & Stinchfield, 2004 ). Gamblers do not necessarily lack the statistical knowledge underpinning many gambling concepts ( Delfabbro, Lahn, & Grabosky, 2006; Lambos & Delfabbro, 2007 ), nor are they absent of insight into the irrationality of their beliefs ( Ellerby & Tunney, 2017; Gaboury & Ladouceur, 1989; Griffiths, 1994 ); suggesting that there is more at play than simply a lack of knowledge regarding mathematical components of probability and chance hindering behavioural changes ( Williams & Connolly, 2006 ). When the rules of probability are disguised within a gambling context, gamblers fail to apply statistical knowledge and instead make decisions based on heuristics and emotion ( Delfabbro et al., 2006; Kogler & Kühberger, 2007; Lambos & Delfabbro, 2007; Turner, Zangeneh, & Littman-Sharp, 2006 ). Since people's beliefs are often value-laden, closely associated with personal experiences, emotions and a sense of self and largely influenced by external factors, interventions that place all the responsibility on the consumer and fail to acknowledge other factors that influence decision making beyond personal control are likely to be ineffective and contribute to gambling stigma ( Carroll, Rodgers, Davidson, & Sims, 2013 ).

Conclusion

Interventions that encourage people to challenge beliefs by providing conflicting evidence, removing blame and stigma, and that explain how and why these justifications are tempting (i.e., the way gambling tricks us into endorsing some beliefs that sway our decisions) are likely to be more effective in promoting cognitive changes that may generate safer gambling consumption amongst those experience gambling related harm. Early cognitive interventions designed to promote greater reflection and challenge biased gambling decisions is likely to encourage safer gambling consumption and have positive implications for treatment seeking by those who need help controlling gambling urges ( Wohl et al., 2007 ). Interventions that are digitalised and can be accessed via the Internet means they can be administered to a wider network of gamblers, and eliminates the stigma or shame associated with accessing formal treatment services. However, a limitation of this study is that it does not delineate the cause for belief and behaviour change. While the intervention appeared successful in altering time spent gambling and gambling cognitions, it is unclear whether the intervention worked by fostering greater analytical thinking that promoted greater reflection more generally, or if it simply promoted reflection on faulty cognitions specifically about gambling; and whether behaviour change occurred as a result of a change in thinking style or due to changes in gambling beliefs. Future research should consider unpacking causation regarding changes to thinking style, gambling beliefs and gambling consumption, and explore the long-term impacts of a training-type intervention designed to strengthen critical thinking skills in gamblers to promote resilience to common gambling related cognitive distortions. Further, exploration as to how the intervention may be adapted or incorporated into other harm reduction strategies to reduce gambling expenditure is necessary to realise the goal of ameliorating gambling related harm.

Funding sources

Part of this research was supported under the Commonwealth Government's Research Training Program/Research Training Scheme. I gratefully acknowledge the financial support provided by the Australian Government and Central Queensland University. Funding agencies have had no involvement in the research design, methodology, conduct, analysis or write-up of this manuscript.

Author's contribution

TG was responsible for the study concept and design, analysis and interpretation of data, statistical analysis, obtaining funding and development of the manuscript. MR contributed to the study concept and design, analysis and interpretation of data, obtaining funding and editing the manuscript. MB contributed to the study concept and design, analysis and interpretation of data, obtaining funding and editing the manuscript. AB contributed to the study concept and design and editing the manuscript for publication.

Conflict of interest

TG has received funding for her doctoral placement from the Australian Government via the RTS program, research funding from the Victorian Responsible Gambling Foundation, and the National Association of Gambling Studies for conference attendance. MR has received research grants from the Queensland Treasury, the Victorian Treasury, the Victorian Responsible Gambling Foundation, the New Zealand Ministry of Health, the NSW Dept of Industry and Trade, the Department of Social Services, the Alberta Gambling Research Institute and Gambling Research Australia. He declares that he has no conflicts of interest in relation to this manuscript. MB received grants from the Victorian Responsible Gambling Foundation, the New Zealand Ministry of Health the NSW Dept of Industry and Trade, the Department of Social Services, the Alberta Gambling Research Institute and Gambling Research Australia. He declares that he has no conflicts of interest in relation to this manuscript. For the period 2015–2020, AB has conducted research funded directly by Australian or international government, or government-related funding agencies, and industry operators. These include Gambling Research Exchange Ontario, ClubsNSW, Dooleys Club Lidcombe, Aristocrat Leisure Industries, Gaming Technologies Association, Gambling Research Australia, Responsible Wagering Australia, Commonwealth Bank, Crown Casino, NSW Department of Trade and Investment (NSW Office of Liquor, Gaming and Racing), La Loterie Romande (Switzerland), Camelot (United Kingdom), La Française des Jeux (France), Loto-Quebec (Canada), and National Lottery (Belgium), Australian Communications and Media Authority’ and the National Association for Gambling Studies. He has received honorariums from Manitoba Gambling Research Program and GambleAware (formerly UK Responsible Gambling Trust) for grant reviews, and royalties from several publishers for books and book chapters. He has also received travel and accommodation expenses from Leagues Clubs, Gambling Research Exchange Ontario, USA National Council on Problem Gambling, Japan Medical Society for Behavioural Addiction, Le Comité d'organisation Congrès international sur les troubles addictifs, Victorian Responsible Gambling Foundation, North American Association of State and Provincial Lotteries, and New Horizons (British Columbia Lottery Corporation to attend conferences and meetings. All professional dealings have been conducted with the aim of enhancing responsible gambling and harm minimisation policies and practices, training counsellors in the treatment interventions, and advancing our understanding of the psychology of gambling.