Shared publicly  - 
A Google+ Experiment in Collective Intelligence
(Post #4 - since the first three posts exceeded comment limits)

UPDATE: I'm closing down this version of the experiment to fix some problems with the below image. Please go here instead:

Please help me run an experiment to test whether the "wisdom of crowds" actually works here on Google+ and see if this community can collectively guess the number of (generic-brand) "Cheerios" in the below glass vase.

The three conditions for a group to be collectively intelligent are diversity, independence, and decentralization. Here are the two steps to run this experiment:

1) Without looking at others' responses (to ensure the independence of our guesses), comment below with your single best guess on how many Cheerios are in the vase. It's one vase, not two; just shown from two different perspectives. Only one guess per person, please.

2) Share this post with your circles to help increase the diversity and decentralization of our guesses.

More info:

Individually, the probability of your guessing the right number of Cheerios is really low, of course. But if there truly is wisdom in a crowd, then when the answers of a large number of people are averaged together, it should converge on the right answer. That's the theory, and that's what we're testing here.

I will let this experiment run until 8PM GMT on April 18. At that point, I'll go through each comment, everywhere it's been shared, plus it to mark that it's been recorded as I input its guess into a spreadsheet. I'll calculate the average and look for some other results and then share our collective guess with you in a follow up post to report on just how wise we are - collectively. To make sure you get it, I will post a comment on every place this post is shared, pointing back to these results.

The ten people with the closest guesses will be featured prominently in that follow up post and I'll also highlight the people who shared the post in which those top ten guesses occurred as a way to recognize at least some of you for helping to ensure the diversity and decentralization of our guesses.

Thanks in advance for your help. I can't wait to see the results and share them with you.

Again, the three conditions for a group to be collectively intelligent are diversity, independence, and decentralization. If you're interested in learning more, here's James Surowiecki's "The Wisdom of Crowds" book on Amazon:

And here's the Wikipedia article on "Wisdom of Crowds":

Edited to note that these are, indeed, not actual "Cheerios" - but a generic brand. Please count the number of "cheerio-like" pieces of cereal. ;-)
Kevin Kobold's profile photoNandan kumar's profile photoDerk JJ's profile photoRichard Chapman's profile photo
So if the number is X, then we will successfully find the answer by everyone just incrementing their guesses one more than the previous one, until we reach X. If we did it that way, I'm not sure it shows anything interesting at all.

I'm joining in with the objectors: this isn't testing collective anything, since these are each independent, random guesses, with no organizational structure tying the guesses together at all.
OK, would you just stop...?




+Daniel Estrada I objected in the first post. This is no measure of collective intelligence. The parameters of the experiment are flawed. If +Gideon Rosenblatt were to ask us to solve for the volume of a single cheerio, and the volume of the vase, and if we could agree or disagree with one another, and stronger arguments could weed out weaker ones, then yes, we would have collective intelligence.
It's got nothing to do with wisdom. It's a statistical experiment: How many guesses does it take to guess the right number, with conditions constrained by Google+ parameters (e.g. user popularity, ripple reach etc).
The answer will be just as boring as the question, and no adding of catch phrases like "wisdom of crowd" will change that.
Assuming those are cheerios and we are not playing by "price is right rules" (otherwise I bet 1), looks like ~845.

*The results of your experiment are being screwed by the ability to see others guess before they guess. If you can find a way to have people guess blind you may get very different (and more accurate) average results.
1414 generic Cheerios in a volume that has a radius of 5 Cheerios and a height of 18 Cheerios.
I didn't guess this time, but let's say 1,465,9273,928,937,609 cheerios are in that vase. +Rahul Roy said 100 million in a previous post, and it made me think. These are guesses, and his one comment skewed the whole thing. There is no collective intelligence here. We are just amassing lots of SEO for Gideon :) And all just to know how many cheerios their are! It's sorta like Facebook, but with cheerios.
Took a cheerio as unit and simulated dropping a whole lot of em a bunch of times in an approximated volume measured with the units.
I also think this is statistical more than anything else, I also think that the end stats should either rule out dumb values or come up with a weighted average to weed out the occasional cheetos filled stadiums... (ex: most guessed ranges should have higher weight in the end result or something.)
(they are the same) because the photos are taken in different points of view.
I am surprised by all the people commenting here who are complaining and have clearly never heard of the infamous Wisdom of the Crowds experiments before. This is exactly the same as those. Look at the wikipedia link and educate yourselves!
+Shem Nyachieo indeed they are, as is stated in the question posed... did you not read it all?

Those guessing more than 500 are idiots or people feeling like this is an optical illusion... the vase is likely 3" up to 7" in diameter if you layered the cheerio pieces flat it would be 18 up to 50 pieces per layer with, my guess of, 15 layers to the top. Also the random angles they sit in the vase which takes up more space... how is anyone guessing over 500.... a hive-mind cant work when half the mind is mindless.
Here is a simple formula for understanding the Wisdom of Crowds theory:

Truth = X + Error

In this case, "Truth" is the actual number of Cheerios. "X" is each individual's guess. "Error" is the difference between one's guess and the actual number of Cheerios.

For example, if there are 500 Cheerios in the bowl ("Truth"), and you guessed 400 ("X"), your "Error" would be +100. If John Doe guesses 600, his "Error" is -100, etc.

In theory (and especially if the conditions of diversity, independence, and decentralization are respected - and people have an incentive NOT to skew the results with absurd guesses such as stated above (i.e. 100 million), when you combine together all the guesses ("X") and average them out (Francis Galton actually considered the "median" to be the best reflection of a crowds estimate, instead of the "mean"), the randomly distributed "Errors" will cancel each other out and all you will be left with is "Truth".

Voilà! It's not magic, it's simple statistics!
The answer = "0", there are no Cheerios in the vase, those are generic oat circles!
i would think based on the width and length in there it woul;d be around 345 to 400
I paged down to end to avoid reading others as I was interested in the outcome of this if the instructions were followed. Unfortunatly when others don't play the game properly or comment with ridiculous answers in order to purposely scew the results the experiment does not work. Unless of course this is actually a social experiment to test peoples reactions to being treated in a conformist manner, if so the results would be interesting.
I think the presumption is wrong. If the chance of an individual's guessing right is very low, it's unlikely that the probability of the crowd's guessing correctly would be significantly higher.
I further believe that people's wisdom cannot be measured by guessing the number of objects. Correct guess of a quantity is simply no wisdom. [Some idiot savants could guess more correctly than an ordinary mathematician!] Wisdom is built on a collective knowledge, based on historical experiences, handed down generations. This experiment CANNOT prove anything of that sort of wisdom.
It's just generic cheerios people... my guess is 863
im gonna go with 0 because those are not cherrios they are generic brand cherrios
I'm thinking 180 in the outermost layer, + 120 + 80 + 60 + 10 = 450
this question test doesnt measure any kind of wisdom/intelligence since you left out some crucial information, namely the dimensions of your glass vase, unless it is a part of the test to googlefu down the specific product and work from there i suppose . . .

as it is presented now though, it's just a massive guesswork~~
I'm going to say 2 cups of Cheerios in each vase. Final answer
Derren Brown did this in U.K and managed to guess the lottery numbers (apparently) so here is my guess - 657
looks like a Piaget conservation task, ;-)
OK, if the top is ~5.5 cherios in diameter and the bottom is about 3.5 cheerios in diameter. And assuming:
80% angle
that 1/3 a cheerio diameter equals the thickness of a cheerio
A 9 cheerio height,
a packing factor of .746 (common toroid packing factor)

you get 355.4 cherios. I'll round up and say 356
i simply counted ten cheerios, and figured how many Ten clusters there are. lol Using my mouse as a guide
Maybe the essential assumption is missing: The individuals are intelligent to begin with.
Liz L
583 but
how does this actually
do anything to prove
collective intelligence
I don't think those are cheerios - they look like some bargain brand version... so I'm gonna say none. (just to be literal)
I'd go with 500ish. My 'method' is countt and multiply. i made it c. 7 accross the bottom; 14 down, and 11 wide = 980. Then I figured that to allow for the roundness and the taper, roughly divide by 2 and round to a round number so as not to look too pretentious!

I'm probably miles out!!!

Interesting experiment, though. I look forward to the results.
He's going to make a histogram (that's like a bar chart) of all these answers and look at where the highest peak is. If the theory about the wisdom of crowds is right, that peak should be close to the right answer even though the guesses vary a lot.

Obtained by:
1. Guesstimating the diameter of a 'cheerio' to be ~1.5cm and the height to be .4cm, the volume of space it takes up is roughly 36pi/160 cm^3.
2. Guesstimating the average diameter of the filled portion of the vase to be 6cm with at height of 14cm, its volume is roughly 126pi cm^3.
3. 126pi / (27pi/160) 126 * 160 / 36 = 560

Also, is it an even # like 140, or odd like 143?
There are no generic brand cheerios in the vase. They are all legit cheerios.
This doesn't appear at first blush to be problem which would lend itself very well to crowd sourcing. The skill of estimating is probably a normal distribution, so a small number of people will be better than everyone else. The Dunning-Krugger Effect suggests that it's likely that most of the people who are not good at estimating, will also be likely to overestimate their ability to estimate. So, by what mechanism will crowd-sourcing dial in an estimate closer to the actual count than, say, any random person's estimate? Not at all clear.

Johnny Ball estimates the number of black cabs in London - Bang Goes the Theory - BBC One
If it works, it will be fascinating to figure out how it worked!
awesome i love this picture and by the way guys there are some other people that actually need food so guys if you are not going to eat anything then throw it away
IT'S OVER 9000!!!
Haha... look it up if you don't get the reference.
and i think about 600
my dad thinks there is about 200
and my mom thinks about 115
Add a comment...