experiments with social media

so i've been curious about second life people's twitter behavior and thought it might be fun to do a few experiments. no no, not talking about evil torturous experiments on defenseless digital animals. and i'm not scraping my social network, getting ready to sell them out to advertisers. i'm just genuinely curious about people's behavior with respect to tweeting and re-tweeting.

the following tests won't be held up by the APA as exemplars of rigorous experimental design. but i hope they'll satisfy some curiosity. also, i should point out that i'm not hiding my intentions by claiming some great reward or a chance to win a bazillion dollars that aren't there. no. i'm offering a few, inexpensive rewards. i'm also paying things in Linden Dollars (L$'s.) this is the game script for second life; it's a lot easier to pay people directly in-world than to mail out checks or deal with PayPal.

experiment 1 : re-tweet this please

my first experiment was this simple tweet:

social experiment. what happens when i ask people to retweet retweet requests in the guise of a social experiment? - please RT!

so i'm basically just asking people to please re-tweet a message. i'm offering no incentives to do so, only the pure joy of participating in my personal experiment. after about eight hours, only two people had re-tweeted this message.

experiment 2 : minor financial reward for re-tweeting

the second experiment was this tweet:

okay. social media experiment 2 : i'll give 50 lindens (each) to the first 10 people who retweet this message.

so i'm changing things up here and providing an incentive. if you re-tweet that message, i'll give you L$ 50. before you get too excited, you should probably note that 50 lindens equates to something around 20 US cents (or US$ 0.20 .) so it's not a really big reward.

i doubt this will be a great shock, but more people re-tweeted this message; after about 4 hours, about 5 people re-tweeted the message. so, it turns out that, at least in my social network, more people will re-tweet a message if there's a personal reward.

another interesting factoid: there was at least one person who re-tweeted this message who i can't see. i believe this means they're protecting their tweets and i'm not following them. people who want to try to get more "twitter buzz" by offering a reward should probably remind folk that if you can't see them, you can't reward them.

experiment 3 : rewarding someone else (a non-profit)

here's my third tweet:

social experiment: if 10 people retweet this message in the next 4 hours, i'll donate L$500 to Bridges for Women, http://sl8.us/gAPHcs4_E

in this case, i'm telling people they'll get no direct reward, but their actions can help out a non-profit. what i didn't really tell people is i planned to donate L$ 500 to Bridges for Women's Second Life presence whether i got the requisite re-tweets or not; it seems like we would be teasing them if we didn't. Also, L$ 500 is about US $2.00, so it's not like we're talking about a huge pile of cash. Still, L$ 500 could be useful if they wanted to spruce up their second life presence a bit with a few inexpensive knick-knacks.

this tweet proved way more popular than the previous two. it took about 2 hours for me to get 10 re-tweets. it's possible, however, this may be because there are just a lot more of my twitter peeps looking at twitter in the early afternoon (pacific time) than in the morning. or it's just that people are more motivated by altruism. or maybe small linden rewards just don't register with people.

experiment 4 : follow me and re-tweet and i'll give you some lindens

with my fourth tweet, i'm getting back to appealing to people's self interest:

social experiment : L$50 to the next 10 new followers who retweet this message. (follow me, then tweet your #secondlife avatar name)

but we're not talking about a lot of cash; remember, L$50 is about US$ 0.20 (20 cents.) also note that i'm telling people to publicly associate their second life and twitter identities. some people may not want to do that. sounds pretty straight-forward, right? yes and no. remember, this is for NEW followers. existing followers, in the immortal words of willy-wonka, "get nothing."

this experiment was crafted in part to avoid the problem with experiment 2 where you couldn't see (or reward) people who were protecting their tweets. you can setup twitter to send you an email message when someone starts following you (in fact, i think that's the default behavior) so you'll always be able to see new followers.

after about an hour, i had 7 re-tweets (though one was from someone whom i think already followed me.) i'm confident it won't take too long to get the last couple new followers.

experiment 5 : lindens for current followers who recruit new followers

the next tweet tries to overcome this last impediment:

social experiment : L$50 for the first 10 existing followers who convince new people to follow me. get new followers to tweet your name.

like previous tweets, we're not talking about a lot of cash. L$50 is about 20 cents. but i was curious if it would affect the outcome.

i actually want to publish this post before people on the east coast start going to sleep, and i just posted this tweet. i'm going to come back with results later with results.

observations

i should probably lead off by reminding people they can observe some of the results themselves. i don't protect my tweets, so you can click on these links to see the individual tweets. if you're using "new twitter," you should see a list of people who have re-tweeted the tweet in question.

• experiment 1 : re-tweet this please
• experiment 2 : minor financial reward for re-tweeting
• experiment 3 : rewarding someone else (a non-profit)
• experiment 4 : follow me and retweet and i'll give you some lindens
• experiment 5 : lindens for current followers who recruit new followers

the second observation i made is that i'm capable of producing some really gawd-aweful social experiments. if i tried submitting these experiments and "results" to a peer-reviewed journal, i'm pretty sure they would be rejected. if i tried to submit them as part of a class in social experimental design, i would be happy to get a D.

"real" experiments start with hypotheses; i didn't do this. they then describe what the testing methodology is going to be; i kinda-sorta did this, but in a haphazard way. and they usually describe how the hypothesis will be confirmed or refuted BEFORE conducting the experiment. i totally didn't do this.

so... observation 2: i'm not really doing a good job at being a social scientist here. but honestly, it's okay. these are anecdotal observations and i never claimed to be a trained social scientist. i know little of how sociological / anthropological experiments are constructed; just enough to be confident i'm doing it wrong here.

the third observation: i think twitter is caching the HTML of individual tweet pages. i noticed that if i used the twitter web page to see who's been re-tweeting things, it seemed to change considerably less often than if i used the API.

more important observations

maybe more important are the following observations:

1. it seems more of my followers are active in the afternoon. yes. this is a completely unscientific statement. i'm conflating time with message content. had i repeated the same message in the morning and afternoon, we could have an "apples to apples" comparison.
2. my followers seem to be somewhat altruistic. pat yourselves on the back, twitter followers, you seem to be eager to give away my hard-earned lindens to charities. seriously though, it took only a couple minutes to get 5 people re-tweet the message in experiment 3 and we finished the challenge with a couple hours to spare.
3. larger sums seem more interesting to people than small sums. when i offered L$50 rewards, uptake was a little slow. slower than when i offered to donate L$500 to a charity. so maybe larger sums motivate people more than smaller sums.
4. different experiments didn't seem to affect each other. after tweeting the L$500 challenge in experiment 3, i waited to see if there would be more interest in experiment 2. i was thinking we might see a couple people looking at the "big tweet" and then notice the others. but no dice.
5. some of my followers think i'm a rube. perhaps they're not aware of just how high-brow embedding roald dahl references in pseudo-scientific observations is. but at least one follower indicated i was behaving like a dork for running these experiments.
6. way more of my followers use the #NewTwitter re-tweet feature than copy and paste messages into new tweets. this shouldn't be a shocker, it's a lot easier on the twitter web page to hit the "re-tweet" link than it is to copy or paste. however, other twitter clients make "re-tweeting with comments" much easier. still, it looks like most people chose the easy "single click re-tweeting" route.

applying these observations

let me start this section by saying you're crazy to use these results as justification for any particular action. it would be nice to say "aha! evidence PROVES larger sums lead to more followers in twitter!" but i just don't see that being an interesting or particular defensible statement based on this data. that being said, the anecdotal data collected might be used to inform a more rigorous set of experiments.

but the data, anecdotal such that it is, might suggest a few "truths."

1. your network may respond better at specific times of the day. i'm going to speculate that for most people, the majority of their human followers are in the same time zone as themselves. tweeting at 3AM is simply going to reach fewer people since fewer people are awake. it might be a cool idea for someone to come up with a tool to see when their twitter followers are most active. actually, i think most twitter analytics packages do this already.
2. if you're going to give money away, people respond better to larger sums. i speculate this is because "more is always better" when it comes to cash. a more interesting way to spin this is that maybe some people just filter out low value commercial tweets. for instance, if someone told me "i'll give you a US$1 gift certificate for amazon" and i happened to be in the middle of an involved task, i may simply choose to ignore it. but if they said, "retweet this within 5 minutes for a US$500 amazon gift certificate," i would probably drop what i was doing and start retweeting. so i'm hypothesizing the existence of a "value point" for each person. rewards below that point are not worth the effort. i would be very interested to see someone come up with a "real" experiment for how to determine where this value point is for users.
3. multiple promotions may have a synergistic effect, but i didn't see that here. i'm going to speculate there are a few people out there who even though they're not interested in a particular promotion, after seeing a it, they'll click on a "more info" or "other promotions" button or look at someone's twitter profile. they may find a promotion or tweet that's more in line with their needs or has a better value proposition for them. i guess what i'm saying is, i hypothesize promotions like the experimental tweets i used here get certain people's attention. while you have their attention, it's easy to convince a few of them to look at other promotions. but i think you have a limited window of opportunity when you have their attention. i spread these tweets out over the course of a full day and had pretty poor luck getting cross-pollenization.

so that's it: my deep thoughts about promotions and twitter. please ping me if you wind up doing "real" research in this field, i'd love to hear about it.

update 1

here's a quick update on the last experiment. so far NO ONE has retweeted the last tweet. this could be due to fatigue (seriously, how long can i keep asking my twitter followers to retweet things.) or it could be because it involves recruiting people to do things. the first four experiments required people simply to press a button. but the last experiment required people to go out, find someone to recruit, and get them to do something.

so maybe we can add one more important observation:

7. you get more responses when tasks are easier to perform. the rewards i'm offering are on the order of L$50 or US$0.20 (not accounting for exchange fees.) maybe it's unreasonable to assume people will (as @shava23 says) "risk reputation for 20 cents." or maybe my followers aren't the recruiting type. or maybe you're not properly motivating existing followers by saying you'll reward new followers. discuss.