Walking to the office earlier this week, I stepped right into oncoming traffic while engrossed in a conversation on the mobile phone. Usually quite mindful of my surroundings, at that point I was so preoccupied that the rules of basic traffic safety were nowhere my mind. By a funny coincidence I arrived at the office and started reading an academic paper with the attention-grabbing title “God is Watching You”. 
This paper by Azim Shariff and Ara Norenzayan takes up where a previous psychology study left off. This previous study  found that college students, when told that the ghost of a dead graduate student was spotted earlier in the testing room, cheated less on a computer test than students who were told nothing. One possible explanation for this result is that the suggestion of being watched by a supernatural being may cause or regulate pro-social behavior such as rule-following.
The “God is Watching You” study intended to investigate how both selfish and pro-social behavior, assessed experimentally through the dictator game , is affected by God concepts. Of course a study like this can’t simply introduce God or religious beliefs since participants may infer the research question and adapt their behavior accordingly. Instead the study subjects had to be primed with God concepts on the sly.
The priming took the form of word games, where subjects unscrambled word groupings to form sentences. Apparently this is a standard method to subconsciously prime subjects in psychology experiments. In the main treatment group, some of these sentences contained target words such as God or spirit. In the control group, no sentence contained these words. After priming, the subjects played a one shot dictator game where the subject divides $10 between themselves and another anonymous player.
In the control group, participants offered an average of $2.60 to their counterpart, while the group that was primed by God concepts offered an average of $4.60. Interestingly, both respondents who later identified as religious and those that identified as not particularly religions (including atheists) offered more money after primed by God concepts.
So apparently thoughts of God, activated without conscious awareness, increased acts of generosity between anonymous strangers. Yet it was not only thoughts of God that induced this generosity. Another randomized group was primed with secular concepts related to justice such as “jury”, “court”, and “contract”. It turns out this group offered $4.40 to the other players – a statistically indistinguishable amount from the group primed by God concepts. Apparently the generous behavior was not due to a belief in a supernatural being watching over you (as perhaps suggested by the dead graduate student experiment) but rather due to a subtle reminder of pro-social notions such as “fairness” and “judgment” that may not always be foremost in our mind when making decisions. And these pro-social notions can be embodied in either religious- or secular-based concepts.
The authors don’t explicitly interpret the results in this light but, after my experiences that morning, I do. Attention is necessarily limited (because it is a scarce resource) and priming is a cognitive manipulation in so far as it brings select issues to the fore. Seen in this way, the results of “God is Watching You” are quite consistent with ideas explored in previous posts on DI. We have written about how survey questions on health can induce the study subjects’ purchase of health insurance months  after interview. We also explored how subjects, preoccupied with the task of remembering a seven-digit number, are much more likely to select chocolate cake over fruit salad from a snack tray  than subjects who need only remember a two-digit number. And, as my friend Berk pointed out, general stress can occupy cognitive resources  that might otherwise be used on more attentive forms of parenting.
It appears that – literally – what is on our mind influences our behavior. And it is not just behavior of ephemera such as whether I should eat pizza or broccoli for lunch but whether I participate in the insurance market and whether, in any given moment, my choices are consistent with my underlying moral beliefs.
If this picture reflects a real phenomenon then it can have implications for the interpretation of impact evaluations – if an information campaign (such as “eating vegetables is good for health”) is measured to affect behavior then it may not be due to the informational content of the campaign but simply the cognitive prompt of receiving the message (“veggies!”). And of course it can have implications for the design of policy – in order to induce honest reporting behavior it may be similarly effective to remind all subjects of the possibility of audit and actually audit less than to not send reminders and audit more widely. Hmmm… sounds like a potential study design…