Thursday the Herald Sun (via GamePolitics) reported on the possibility of lie detection games being supported by the new Wii Vitality Sensor. While I’ve not seen any reports that the Vitality sensor measures skin conductance (i.e. level of sweat on the inner surface of the fingers) as claimed in the article*, it did get me thinking whether or not lie detection could be a fun game mechanic.
Basics of Lie Detection
Lie detection is based on the premise that when a person lies it elicits a physiological response which can be discriminated from the truth. In a polygraph test (a type of lie detection test) this premise is used to ascertain whether a person is answering a question truthfully or not using a range of autonomic measures such as pulse rate (i.e. like Vitality supports), skin conductance and blood pressure. In a typical polygraph test an investigator begins by asking a subject a few sample questions for which the truth is already known. This allows them to build a baseline for physiological activity representative of a question answered truthfully. Next the investigator will ask questions for which the truth is not known and via their physiological responses the investigator will infer whether they have lied or not. Obviously this all assumes that lying has its own physiological discriminates. To my knowledge it doesn’t, least not the autonomic measures commonly used in a polygraph test. For more information I suggest consulting The Polygraph and Lie Detection (National Academy Press)**.
Anyways back to whether lie detection can be a fun game mechanic. I’m going to walk you through the design of an example lie detection game and discuss the various issues and its potential for play along the way.
Game Format for Lie Detection
In a lie detection test the investigator’s aim is to verify whether or not the subject’s answers are truthful or not using their physiological responses. The subject’s aim is to hide information from the investigator by controlling their physiological responses and/or deploying countermeasures to create false positives. In a game we would therefore have two players: one seeking information (seeker) and another hiding information (secret-keeper). Assuming we don’t want player’s falling out by letting them ask any old question I’m going to use a game format suggested by my WaveRider Mind Games book***: ‘guess the number I’m thinking of’. In this game the seeker has to guess which number between 1 and 10 the secret-keeper is thinking of by calling out the numbers in order and observing their responses to each call out. This format is apt for play for two reasons: 1) an emotional response is not provoked by the seekers questions therefore limiting physiological interference from other processes (e.g. if the seeker was allowed to ask an embarrassing or other socially awkward question the physiological response will be contaminated with the emotional response providing a false positive) and 2) player’s can’t ask questions someone really doesn’t want to answer. In the later case if the seeker invests too much stock in the response to an emotive question the experience becomes awkward and has implications outside the game. Hence the game is better off using number guessing.
This particular format is the opposite of your typical lie detection test as the hidden information is the truth rather than the lie. The seeker is therefore monitoring their opponent for a response to their recognition of their chosen number when it is called. This makes the game a little easier for the seeker to infer the hidden number than with a standard lie detection test format (i.e. not exactly watching for a more dubious deception response).
A Wii Implementation
In a Wii implementation we can play our game in three different configurations: 1) 2 human players in the same room (format similar to a polygraph test), 2) 2 human players across a network and 3) 1 human player and a computer opponent. In the later configuration the computer would always act as the seeker. Before play begins the secret-keeper selects a number between 1 and 10 to hide. To avoid the secret-keeper from lying about their choice (a valid concern in a lie detection game) the player would enter their choice into the game for the computer to reveal later. The secret-keeper then relaxes so their physiological signals stabilise and then the game begins. The seeker then calls out the numbers in the allotted range and watches for differences in their opponents physiological response. Once they’re happy they’ve identified the hidden information they input the answer into the game and the computer confirms whether or not they were successful. In the network configuration, the game could call out the numbers instead of the seeker.
During play the secret-keeper is not allowed to see their own physiological responses as it helps them deploy counter measures (e.g. if they can see they responded badly when the hidden information was queried they can deploy a countermeasure to throw the seeker off track on the next query). Therefore in the first game configuration only the seeker can see the game screen whereas in the second and third configurations the secret-keeper would see anything but their physiology.
The challenge for the seeker is pretty straightforward: identify which number their opponent is hiding using their physiological response to number call outs. If the secret-keeper can not inhibit their response when the hidden information is called the seekers wins.
The challenge for the secret-keeper is much more interesting and is two fold: 1) self-regulation of their physiological state to mask any recognition of a call out of the hidden information and 2) deploying counter measures. Counter measures are used to throw the seeker off-track with false positives, and are done by voluntary provoking a physiological response. For example if the physiological measure being used to infer the secret-keeper’s state is skin conductance, biting the lip will cause a noticeable increase in conductance to throw them off track as the seeker will only expect an increase when their opponent recognises the call out. The secret-keepers counter measure strategy can take a variety of forms and its up to the seeker to identify and ignore them to discern which number their opponent is hiding.
The Vitality sensor measures the players pulse rate (aka heartbeat rate) as derived from changes in blood volume as the heart pumps blood around the body. For an informal setting such as a game, this signal is not the most suited for play as pulse rate is predominately influenced by a persons breathing which changes the signal in a cyclic fashion to the inhalation and expulsion of air (nb. you take about a dozen or so breathes a minute at rest). As such the seeker will require more time to identify meaningful changes as they have to account for the breathing cycle. Because of this the game becomes more difficult to play for the seeker as they have to spend more time interpreting their opponents physiological responses. Abstractions of pulse rate or other derivatives could be used however again this makes the game longer and is not very entertaining if your the secret-keeper who essentially sits still for most of the game. A more suitable measure would be skin conductance which increases when a person is aroused and falls back to a baseline when not thereby making it relatively easy to infer a response to a call out. Also the latency on a response is of the order of a few seconds, which allows for repeated queries in short succession unlike the use of pulse rate which will be longer (less the seeker is aided by the computer to help interpret the data).
In practise lie detection involves a lot of repetition in order to determine if a response giving away the hidden number is geniue or not. This increases the number of times the number range has to be called out. We can reduce this by enforcing a time limit on the seeker and/or restricting the number range the secret-keeper can select from, however this increases the seekers chance of guessing the hidden number by chance.
Excessive counter measures
The more excessive the counter measure strategy the longer the game is drawn out as all responses become contaminated with noise. If too many counter-measures are deployed the game becomes unplayable and so is reliant on the secret-keeper deploying counter measures in a limited fashion (e.g. generate noise on a specific call out other than the hidden number to confuse the seeker).
Overt vs Covert Responses
In the first of the game configurations: two players in the same room, the seeker can use their opponents overt responses (e.g. facial expressions) to identify the hidden number rather than by the provided covert response which becomes redundant.
Informal vs. Formal Lie Detection
The Wii is predominately aimed at a casual audience and based on Nintendo’s choice in physiological signal exegaming / and relaxation games appear to be their inital market. Lie detection relies on a formal setting for a stable physiological signal in order to discriminate between responses. An informal setting as with an entertainment product is liable to generate a lot of physiological interference, and unlike a lot of relaxation game where interference feeds into the game mechanic, it is not a desired setting for a lie detection game. As such lie detection is not a suitable game format without the necessary rules to enforce serious behaviour, disturbances like laughter will intefer and require a cool-down period to stabilise the physiological signals
Conclusions and Other Interesting Applications
As a game in itself, lie detection suffers from numerous issues which make it unsuitable for play, especially the market Nintendo is aiming for. I could of taken a more light hearted approach to lie detection and of allowed players to ask any question they fancy however the secret-keeper’s responses become a lot more dubious to interpret as belonging to the truth of their answer. I imagine the game’s novelty would wear off pretty quick when the physiological response could rarely be relied upon. However the more controlled approach I’ve taken to a lie detection game makes it more of a chore than an entertaining experience.
However that is not to say we could not use lie detection in different formats. For example in a poker game, being able to control one’s overt responses is part of the game experience. So in a networked Wii game, we could change the player’s avatar according to their covert responses (e.g. if their baseline heartbeat rate increases over a minute change the colour of their avatars clothes). If each round of poker lasts a few minutes players can easily make use of these changes to make decisions in the game in a similar fashion to how overt responses would be used. The challenge element for player’s comes from making reasonable inference from their opponents physiological changes to the current game context. It’ll be interesting to see what Nintendo and 3rd party publishers come up with if they attempt to use this mechanic, for the time being I’d be a little wary of its fun quality especially having played a few rounds with WaveRider.
If anyone else has any thoughts on lie detection as a game mechanic drop us a comment below.
As an aside, it’ll be interesting to see how live streams of the player’s physiological state will impact social interaction during play. I once used MSM’s ‘What are you listening to’ feature to provide my contacts with live heartbeat rate data (among other derivates like heartbeat variability) to see how they’d respond. Unfortunately a few minutes after switching the thing on one friend in particular spent his entire time hurling various insults to see if he could get my heartbeat to rise.
* For skin conductance two metal contacts are required to push an electrical current through and thereby measure skin resistance; conductance is the reciprocal. These contacts are typically placed on adjacent fingers. The available production shots only reveal a single finger-based-sensor and the small square plate seen inside is clearly the pulse oximeter sensor. Nintendo’s press release on the sensor isn’t too clear whether they intend to add new sensors (e.g. as additional finger-based-sensors making the device similar to the LightStone/IOM) or they intend to use other physiological derivatives from the pulse oximeter (e.g. heart rate variability).
** This report reviews the theory behind lie detection and polygraph tests as well as existing studies; essentially I wouldn’t take much stock in the accuracy of lie detection (at least not lie detection based on autonomic measures as with polygraphs).
*** WaveRider is a biofeedback device I’ve been using for experimental affective games for several years.
I like the idea of lie or truth detection being used in gameplay context. From my reading of the literature, it may be possible to get something from heart rate. Phasic (short-term) changes in heart period (expressed in ms) do respond to cognitive challenges like mental arithmetic. As I remember, heart period increased (heart rate slows) during internal processing and increases during preparation for action. You’d have to look at heart period for 4-8 beats before and after the question from the seeker to see anything, but it’s possible.
I also wondered about people playing it using mobile devices in the field (which would make the game more challenging for the seeker).Inevitably, people like this concept because it’s about accessing private, secret thoughts and preferences. My gut feeling is that versions would soon appear where so-called friends bombarded one another with dodgy pictures or film footage just to get a response (as your mate did).
I think online_poker is the killer application for this concept. Basically, the elements of disguising your own expression and reading the behaviour of others is already part of the game mechanic. Adding physiological feedback from each player adds a new dimension to gameplay.
The site doesn’t like the term ‘online_poker’ (separated using either a whitespace or hyphen); that’s what was triggering the ‘412 Precondition Failed’ error. Use the term ‘network poker’ or something similar until I figure out a fix.
Just reviewed an abstract for a human factors conference where EEG was used to discriminate between ‘bluffing’ and ‘giving up’ during human-human game of poker. I will follow up on that later in the year when the work is presented.
Pingback: Physiological Computing : Predictions for 2011
Pingback: Physiological Computing : CHI 2011 Workshop – NASA does Biofeedback Gaming on the Wii