Opendata, web and dolomites

Report

Teaser, summary, work performed and final results

Periodic Reporting for period 1 - SmartMass (Behavioral and cognitive mechanisms promoting collective intelligence in human groups)

Teaser

We want to identify the mechanisms by which individual behaviors can promote the emergence of collective intelligence in human groups. We want to know how the information sharing between individuals can lead to efficient collective decisions (micro to macro), and to identify...

Summary

We want to identify the mechanisms by which individual behaviors can promote the emergence of collective intelligence in human groups. We want to know how the information sharing between individuals can lead to efficient collective decisions (micro to macro), and to identify the specific situations where individuals behaviors and choices are affected by the group (macro to micro).

The importance for society is in determining if collective intelligence arise because of the intrinsic nature of individual behavioral rules, or because individuals are able to identify the information required to coordinate their actions and make efficient decisions, which factors contribute to a successful information process, and what are the effects of insufficient information, information overload, and bad quality information.

By combining insights and methods from diverse disciplines, we want to determine 1) the basic interactions between individuals and with their environment, in particular, the kind of information extracted and used from the environment and exchanged with other individuals, and the mechanisms of transmission and propagation of this information among individuals; 2) the effects of these interactions on individual choices and actions, 3) the resulting collective behavioral effects, and 4) the influence of group behavior on individual interactions.

The SMARTMASS project has successfully created a hub for interdisciplinary research, centered on addressing behavioral and economic problems with a methodology combining experiments and mathematical modeling. New ideas have been unlocked by associating experimental and theoretical tools of cognitive sciences, quantitative ethology, statistical physics, economics and game theory.

We have designed and executed specific experiments to investigate how human groups select alternative solutions to solve complex problems under different conditions of information. We have proposed different scenarios, derived mathematical models from experimental data, and studied these models by means of classical and novel tools.

Main conclusion:
Human groups are very effective in solving complex problems collectively. The accurate delivery (in quantity and quality) of local individual information definitively allows to enhance collective intelligence in human groups.

Work performed

\"The work consisted on the design, execution and analysis of five tasks aimed to reveal the mechanisms by which collective decisions arise from the processing and interchange of information between individuals.
1) Optimizing individual decisions within a group: A group is confronted to several possible choices in estimation tasks under the influence of other individuals\' choices and the advice of experts. The experiment allows to quantify how information delivered to individuals affects their choices, and to determine their ability to retain or discard information.
2) Collective decisions in pedestrian crowds: How mimetic behavior, interactions between individuals, information sources and past experience affect the decisions of pedestrians walking in a corridor with bifurcations? Subjects holding a helmet of VR are asked to choose the passage they estimate to has been the less often chosen according to their own perception and partial information provided by panels.
3) Information processing in segregation of pedestrians: What is the average number of neighbors that provides the optimal amount of information of pedestrian\'s environment enhancing the performance of the group in a collective task?
4) Characterization of human walk and pedestrians\' interactions in a confined space.
5) Collective search for information: Identification of individual strategies based on past social information in a searching task. Subjects have to find the maximum value of the global quality of a product by exploring a square grid of NxN cells.

We have implemented the experimental setups for each task, carried out the experiments, the data analysis, the formulation of mathematical models and their analysis, tested models\' predictions, and carried out new experiments, according to our research methodology.

Results:
- Social influence can help groups to improve their performance and accuracy in estimation tasks. Group performance can be enhanced by delivering incorrect information to the subjects. Even with little prior knowledge, individuals are able to use peers\' information and collectively improve group performance. Getting a better understanding of these influential processes opens new perspectives to develop information systems aimed at enhancing cooperation and collaboration in human groups.
- The introduction of Virtual Reality as a fundamental tool for scientific research. VR makes possible exact replicability and instantaneous modification of the experimental environment.
- More than 800 hours of human walk in different conditions. Analytical expressions of the interaction functions of pedestrians with their environment. Humans in crowded conditions are impressively efficient in solving complex tasks at the collective level with a very limited amount of information. Our models may guide the design of pedestrian pathways and spaces that should facilitate pedestrian flows.
- A multi-platform game to identify individual and collective strategies of searching information. An experimental protocol that will provide insights about individuals\' information processing in five starts evaluation systems (amazon, tripadvisor), especially in discarting false information.

Two workshops:
\"\"Can we enhance collective intelligence in human groups?\"\", 14-15 Apr. 2016 (TSE)
\"\"Analyzing and modeling individual and collective behavior\"\", 07 Dec. 2016 (CIB), Toulouse.
\"

Final results

\"Progress is found in the conception of the core issues we posed, the experimental protocols, the data, our methodology, and the results.
In particular:

- accurate characterization of \"\"social influence\"\" and \"\"social information\"\" in the context of collective estimation,
- noticeable prediction that incorrect information can lead to better collective decisions, which could have a strong socioeconomic impact in decision making strategies in governments, policy makers and companies,
- physical and cognitive characterization of human walk and interaction with the environment and other pedestrians, including the strategies of decision making in binary choices with partial information, which are of crucial interest for the optimal design of both the architecture of public spaces for pedestrian circulation and the control in real time of pedestrian flows, crowed or not, evolving in conditions of normality.

Methodological progress can be found:
- in our experimental procedures, where the amount and quality of information provided to- and interchanged by- subjects are controlled
- in the introduction of Virtual Reality as a fundamental tool for fundamental research, providing experimental environments where exact replicability and immediate adaptation of the environment to the instantaneous behavior of subjects are possible
- in the use and fine interpretation of Ubisense, a tracking system in real time for up to 22 persons with a frequency of 2Hz and a precision of 20-30 centimeters
- in the development of a multi-platform application for collective search that can display several environments of decision making and social choice such as the ubiquitous 5-stars systems of evaluation in social networks (amazon, tripadvisor, etc.).
\"

Website & more info

More info: http://cognition.ups-tlse.fr/spip.php.