A new way to increase the wisdom of the crowd
Author: Sergey Karelov
Source (the text has been edited)
The new study of five universities in the US and Germany, together with two institutes (Santa Fe and Max Planck) - this is the next, the fifth in the past two years, a breakthrough in technology to enhance collective intelligence.
The wisdom of the crowd (a phenomenon that was known and described a long time ago and became terribly popular after the book of the same name by Surowiecki) lies at the core of crowdsourcing - the most developed class of technologies for building and exploiting collective intelligence for decision-making.
The last 2 years have become critical for crowdsourcing. There were published four breakthrough studies that could revolutionize crowdsourcing and, dramatically increasing its effectiveness, turn collective intelligence into an acknowledged mechanism for making complex decisions in the 21st century. But there was not enough the last link - experimental confirmation of the mother model as it was before 1919, when the gravitational deviation of light predicted by Einstein was confirmed.
And here is the last link, finally, in the fifth breakthrough study of crowdsourcing, aimed at the eradication of three inherent fundamental shortcomings:
1) information noisiness (smart thoughts, ideas and utterances sink under thousands of simple and banal utterances);
2) low efficiency of the collective work (the final intellectual level of the crowd is easily lowered to the level of its "weakest" participants);
3) inclination of the majority to join the most popular opinion.
The key reasons for all three named congenital ailments of classic crowdsourcing, one way or another, are (singly and collectively):
А) individual bias;
B) social influence.
The search for effective ways to deal with 1-3 (and correspondingly with A and B) is the main task of modern research on increasing the effectiveness of collective non-expert (mass!) Intelligence - "wisdom of the crowd" and crowdsourcing, as a way of extracting and integrating this wisdom.
As for the new research:
✔️ For the first time it was possible to measure both factors (A and B).
✔️ It has proposed corrective amendments, significantly reducing the impact of both factors.
Imagine an archer who perfectly knows how to shoot at a dash on static targets, but who can not take into account when shooting in the open area the distance and speed of the target's movement. Today's classic crowdsourcing is a shooter who learned to shoot at a dash and smear when shooting in the field on live rabbits, and a new study gives the arrow a table of corrections allowing you to take into account the distance and speed of the hare in the course of aiming. The result is remarkable. Accuracy of shooting is much higher.
The authors of the research proposed 3 groups of amendments to the aggregate estimates of crowdsorbers:
— correction of mean value (averaging all estimates);
— median correction;
— correction to maximize the probability of a correct estimate.
Two Important Remarks
1. The proposed method of making corrections is applicable so far only for numerical estimates made by crowdsorsers (for example, estimate by eye how many peas in this bank). And this is only a very limited range of tasks. Where more often crowdsorbers require non-numerical estimates or predictions. However, the method is very promising, and in time it will be tried to transfer to non-numerical estimates.
2. Instrumental methods of leveling individual bias and group influence during crowdsourcing began to be studied in the West only in the mid-2010s.
At the same time, in Russia, specifically to solve the problems of leveling individual bias and group influence, back in the late 2000s, an original method was developed, called (Syn)intellectual crowdsourcing. Moreover, this new, much more effective method of crowdsourcing in the early 2010s was implemented in the software platform Witology, which was executed under a hundred commercial projects.
The individual prejudice of crowdsourcers, when working on the Witology platform, is identified and minimized with the help of an advanced rating system of participants, which determines the "weight of opinion" of each of them in the integrated "crowd opinion". Well, social influence in the development of their own opinion crowdsource, working on the platform of Witology, generally reduced to zero. The participant's user interface does not allow him to see either the popularity of other people's opinions, or even these opinions. They become available for public viewing only after fixing their own opinion by each of the participants. As a result, the "Law of Matthew" on the platform of Witology does not work.
To me, as the author of the synintellectual crowdsourcing, of course, this rather rare example of at least temporary superiority of Russian commercial IT technology over the western one is pleasant...
— 1. A radical change in the basic algorithm of crowdsourcing (two questions instead of one).
— 2. The change of crowdsourcers' motivation (reward for "minority report").
— 3. The method of isolation in the "crowd" of the so-called an "informed minority".
— 4. Improving the quality of collective decisions by increasing the proportion of "stubborn dissenters".