The use of deep learning technologies in massive citizen consultation processes allows to quickly and efficiently process thousands of proposals in order to give viability to this type of democratic exercise.
Presidential elections are being held in Chile. At least six candidates are contesting this election. One of them, the leader in the polls, Gabriel Boric, decided to create his government program from an unprecedented process, in which hundreds of citizen groups were created to discuss ideas and proposals to be included in the government program in dozens of priority areas.
Between September 20th and October 24th, 2021, a citizen programmatic construction process was carried out by Gabriel Boric in which 33,728 people participated both in the form of Citizen Tables as well as via online participation.
A total of 603 Citizen Tables for Boric were organized in which a total of 7,284 people worked in 80 communes of Chile, in addition to 14 tables in foreign territory. Of the total number of Citizen Tables, 258 correspond to Territorial Tables and 318 to Technical-Thematic or Citizen Cause Tables, and 27 Sectorial Work Tables.
In the Online Citizen Consultation, more than 120 thousand responses were submitted by 12,011 people who participated in the online process, while 16,433 people responded to the citizen questions posted on social networks. As a result of the process, a total of 13,250 programmatic proposals were raised.
Below is an infographic summarizing the process:
More than 13,000 programmatic proposals received had to be processed in a few days in order to present the result of this process to the country. For this purpose, and given the impossibility of doing it manually, natural language models were used as a Deep Talk service. .csv files with the proposals of individuals and programmatic tables were uploaded to the online platform, and from this, 90 clusters were obtained around which the proposals were organized. Here are the clusters (Spanish):
Then a more detailed analysis of these clusters was made to group some similar ones and their proposals with which a summary of suggestions according to areas was built.
The use of deep learning technologies in massive citizen consultation processes allows to quickly and efficiently process thousands of proposals in order to give viability to this type of democratic exercise. In the case described above, the amount of human resources required to manually process the large number of proposals written in natural language made its implementation very difficult. With the use of the Deep Talk platform, thousands of proposals could be processed in a matter of minutes, providing real viability to the process.
It is interesting to observe how the use of this type of tools can support the deepening of democracy through the active participation of citizens, writing, giving their opinions, and generating programmatic proposals to governments.
Official Document here