The production and detection of deception in an interactive game
Research highlights
▶ This study examines production and detection of deception in a two-person game. ▶ The study allows for free choice in decision-making. ▶ Deception is not always equivalent to telling a falsehood. ▶ Deception is facilitated by hierarchical decision-making and risk evaluation.
Introduction
The question of how people deceive and how they detect deception has been widely discussed in the literature of psychology, forensics, and recently, neuroimaging (Ganis et al., 2003, Kozel et al., 2005, Langleben et al., 2005, Spence et al., 2008). As is widely recognized (e.g. Greely and Illes, 2007, Sip et al., 2008a, Sip et al., 2008b; see also Abe et al., 2007, Greene and Paxton, 2009), one of the main questions for experimental studies of deception is the degree to which laboratory tasks can elicit the same processes as those occurring in ‘real life’ settings. However, the closer we come to a real life setting in the laboratory, the more we must sacrifice tight experimental control.
In one of the first studies attempting to address the issue of ‘instructed lies’, Abe and colleagues (2007) introduced a very elegant twist to their experimental protocol. The twist takes place at the level of the instructions, where the participants received two mutually contradictory instructions from two experimenters. In practice, during the break necessary for the radioactive tracer to decay, when experimenter 1 was not present in the room, experimenter 2 secretly asked the participants to deceive experimenter 1 by answering questions with opposite responses than those suggested by the experimenter 1. The main effect of deceiving one of the experimenters, that is ‘falsifying’ otherwise truthful responses, showed increased neural activation of the left dorsolateral and right anterior prefrontal cortices. This study, though it does resonate with real-life deception, still suffers from a problematic issue of ‘instructed lies’. In this scenario, the participants faced an externally introduced change to suddenly obey a different set of rules. It is difficult then to account for that change in terms of neural activation in reference to the processing of deception as the actual peripheral attentional activation and deception activation could have contributed to the final results. Therefore, to address these issues in our investigation and to avoid instructing the participants to ‘lie’ at any stage (see also Baumgartner et al., 2009, Greene and Paxton, 2009), we chose a paradigm which fully allows the participants to be deceptive at will.
As a compromise between ecological validity and experimental control, we have developed a laboratory version of the deception game ‘Meyer’, widely played in Denmark, in a form suitable for the scanning environment. One advantage of using this game is that participants must decide for themselves when to deceive, a decision based on the current state of the game (see below for a detailed account of this game). This is in contrast to the many studies of deception in which participants are instructed by the experimenter exactly when to be deceptive, that is to tell a falsehood (see e.g. Ganis et al., 2003, Langleben et al., 2002, Langleben et al., 2005, Mohamed et al., 2006). Another advantage is, that, by using a two-person game, we can look at the detection as well as the production of deception. The corresponding disadvantages with the game include loss of control over how many deceptive events will be generated and, in the detection condition, the participants’ expectation that deception will occur.
Our use of a two-person game emphasizes the inherent communicative nature of deception. We characterize deception as an intentional attempt conducted by an agent (A) to induce or reinforce in another individual (recipient) a belief that, in A's opinion, is false (Sip, 2009). One implication of this characterization is that deception is interpersonal and interactive; another is that being deceptive is not synonymous with telling a falsehood (see also Abe et al., 2007, Greene and Paxton, 2009). The agent may intentionally create the belief in the recipient that he is ‘lying’ when he is actually telling the truth. Such intentions are an important feature in the game of Meyer in which two players take it in turns to throw two dice. The players may attempt to accomplish deception by telling the truth with a deceptive intention. The players are effectively exchanging statements, referred to in this paper as claims. Players in turn make claims about the state of the dice (production moves), and respond with either acceptance or rejection of the other player's claim (responding moves). The game progresses much in the manner of a conversation, with interlocked turns, or moves. A player's claims are not isolated responses, but are tied tightly into the social context and history of the game in which they occur. Players must take the intentions of the claimant into account, requiring inferencing informed by the interpersonal and interactive nature of the game.
Although many different paradigms have been used in neuroimaging studies of deception (for reviews see Greely and Illes, 2007, Sip et al., 2008a), there is a consistency in the pattern of brain regions activated (Christ, Van Essen, Watson, Brubaker, & McDermott, 2009): anterior cingulate (ACC), dorsolateral prefrontal cortex (DLPFC), medial prefrontal cortex (MPFC), and inferior frontal gyrus (IFG). However, these regions are clearly not dedicated solely to deceptive behaviour. Rather, deception draws on a number of cognitive processes that are not uniquely concerned with deception. For successful deception, there is a need to plan ahead, to inhibit prepotent responses (i.e. to tell the truth), and to keep track of the opponent's beliefs, all of which processes are associated with activity in the prefrontal cortex (e.g. Amodio and Frith, 2006, Koechlin and Summerfield, 2007, Shallice, 1988).
Recently, Koechlin and colleagues (Koechlin et al., 2003, Koechlin and Summerfield, 2007, Koechlin and Hyafil, 2007, for review see Ramnani & Owen, 2004) have developed a hierarchical model of cognitive control of executive brain functions in action selection. In the latest version (Koechlin & Summerfield, 2007), the model effectively identifies four levels of cognitive control of action with respect to earlier events. The model defines the process of action selection as a “fractionation of cognitive control according to the temporal framing of actions and events involved in selection” (Koechlin & Summerfield, 2007, see also Braver & Bongiolatti, 2002). The first and lowest level is a sensory control in selecting motor action, with activity in premotor regions. The second level is associated with a contextual control involved in responding to stimuli in the immediate external context in which these occur with activity located in caudal LPFC (typically BA9/44/45). The third level is an episodic control with specific activity typically in DLPFC (BA46). Finally, the fourth level, a subdivision of the episodic control, is a branching control located in the lateral fronto-polar cortex (BA10).
We would expect these levels to be present in decision-making in Meyer, however, in contrast with many other paradigms aimed at studying deception, the most prominent aspect of action selection in this study lies in the branching control. This is due to the fact that the episodic control is assumed to be divided into a control of signals, both subsequent and preceding ones (Koechlin & Summerfield, 2007). As the branching control appears to allow for accommodating a pending choice to an ongoing behaviour, it significantly contributes to the construction of a coherent deceptive context which accounts for past and upcoming events. In the game of Meyer it is of particular importance for the player to decide whether or not to be deceptive after each dice throw in relation to past choices and the stage of the game.
There is a difference between true and false claims, specific to Meyer, which is not directly related to deception. When making a true claim the player reports the value of the dice throw. When making a false claim, the player must choose which claim to make. Thus, when making a false claim, additional processes will be engaged concerned with selecting responses from among competing alternatives. This process is likely to involve a frontal-parietal network (see Schumacher, Cole, & D’Esposito, 2007 for a review).
Interestingly, much less previous work was done on detecting deception (responding moves in Meyer) (see e.g. Grezes, Frith, & Passingham, 2004). In Meyer, the player has to decide whether the opponent is being truthful. If the player accepts the opponents claim there is no immediate consequence, since the game continues. However, challenging the opponent's claim is a more risky strategy with an uncertain outcome, since the player may lose.
Section snippets
Stimuli
Participants were examined in a 3 T scanner while playing a computerized version of Meyer with a human opponent outside of the scanner (confederate). The participant and the opponent played in turn, as in the standard game of Meyer. However, the data collected during the times that opponent was making his choices were not included in the analysis. The game was divided into 4 game sessions of 5 min each and one 5 min control session. The game sessions involved each of the 4 possible combinations of
Behavioural results and reports
Since the experiment employed a real-life paradigm, neither the number of true and false trials nor their placement was controlled. Games consisted of an average of 35.93 ± 0.97 production moves and 34.43 ± 0.56 responding moves. On average, in production moves participants claimed truthfully 21.1 ± 1.1 times, and falsely 14.1 ± 0.9 times. In responding moves they rejected their opponent's claim on average 15.6 ± 0.34 times, and accepted it 18.9 ± 0.6 times. In the control session, participants claimed
Imaging results: production moves
As already mentioned, in the control session, in production moves the subject simply reported the current throw of the dice, whereas in production moves in the game sessions participants could claim either falsely or truthfully. In contrast to the control condition, claiming falsely in the game session was associated with activity in a network of frontal and parietal regions (see Table 1 and Fig. 4).
In comparison to the control condition, claiming truthfully activated a subset of these regions,
Imaging results: responding moves
Surprisingly, the comparison between rejecting and accepting the opponent's claim does not survive the corrected threshold (see Supplementary Materials for Details). Unexpectedly, there was a significant decrease of activity in the posterior STS when claims were accepted or rejected in game sessions in comparison to control sessions (see Fig. 5).
Discussion
In the current study, we presented a deceptive dice game in which players expect their opponents to attempt to deceive them, and are perhaps more alert to the possibility of deception than in casual conversation. This expectation is exploited by players, who make true claims, but do so in ways designed to invoke the inference that they are making a false one. In the context of Meyer, a true claim is often used with as much deceptive intent as a false one. An advantage of this strategy is that
Conclusions
Our study of deceptive behaviour in a popular game suggests that the neural correlates of deception depend very much on context. In this game claiming truthfully can be deliberately deceptive, and activates similar areas to claiming falsely. Our results are in accordance with Koechlin's proposed three level hierarchy of executive control (Koechlin et al., 2003, Koechlin and Summerfield, 2007), and suggest the following relationships: (1) activity in BA10 is associated with setting a deceptive
Acknowledgements
This work was supported by The Faculty of Humanities at the University of Aarhus, The Danish Research Council for Culture and Communication, and The Danish National Research Foundation's grant to Center of Functionally Integrative Neuroscience.
References (37)
- et al.
The neural circuitry of a broken promise
Neuron
(2009) - et al.
The role of frontopolar cortex in subgoal processing during working memory
Neuroimage
(2002) - et al.
The gateway hypothesis of rostral prefrontal cortex (area 10) function
Trends in Cognitive Science
(2007) - et al.
Analysis of fMRI time-series revisited
Neuroimage
(1995) - et al.
Imaging the intentional stance in a competitive game
Neuroimage
(2002) - et al.
An information theoretical approach to prefrontal executive function
Trends in Cognitive Sciences
(2007) - et al.
Detecting deception using functional magnetic resonance imaging
Biological Psychiatry
(2005) - et al.
Brain activity during simulated deception: An event-related functional magnetic resonance study
Neuroimage
(2002) - et al.
The neural correlates of theory of mind within interpersonal interactions
Neuroimage
(2004) - et al.
People thinking about thinking people. the role of the temporo-parietal junction in “theory of mind”
Neuroimage
(2003)