Skip to content
Home » Breaking News Tokyo gubernatorial election 2024, AI tests and scores the claims of each candidate

Breaking News Tokyo gubernatorial election 2024, AI tests and scores the claims of each candidate

[Breaking News] Tokyo gubernatorial election 2024, AI tests and scores the claims of each candidate
Press release: July 6, 2024
[Breaking News] Tokyo gubernatorial election 2024, AI tests and scores the claims of each candidate
*As a result of scoring the political opinion broadcasts of 2024 candidates based on the number of votes received in 2016 and 2020, the number one political opinion broadcast with a tendency to get a high number of votes was Shinji Ishimaru, and Mr. Koike’s opinion diverged from the previous comparison*
Shinji Ishimaru ranked first in political broadcasting with a tendency to get a high number of votes, while Koike’s opinions diverged in the previous comparison.
Cognity Co., Ltd. (Headquarters: Shinagawa-ku, Tokyo, Representative Director: Rie Kono), which conducts AI tests for communication such as business negotiations and leadership skills.
Cognity) uses AI to test the political broadcasts of Tokyo
gubernatorial election candidates in 2016, 2020, and 2024, and the survey results “Captured with AI talk test.
Tokyo gubernatorial election 2024” announced.
In this survey, we used AI to examine the political opinion broadcasts of all 86 candidates from the past three Tokyo gubernatorial elections, and explained the trends of the candidates with the highest number of votes and the characteristics of this year’s candidates.

Click here for details:

* *
* ■Extract the main arguments of each candidate in ranking format* This time, we examined the political opinion broadcasts of 48 people, and extracted in ranking format the topics that were well-argued, had a large amount of explanation, and included evidence.
[Testing method] Patented technology CogStructure
, to quantify what was said and extract claims. Explanatory
information for various lengths is classified into logical information and repetitive/specific information. Based on the “argument structure that is easy for people to understand and accept” derived from 60,000 talk data accumulated in the past, the target arguments are scored and presented in the order of their ranking.
Ranking of each candidate’s main arguments
* -Claim detection case: Mr. Renho- *
[1st place] We will create a glass-walled metropolitan government and connect it to new business opportunities.
[2nd place] Thoroughly support young people
[3rd place] Tokyo, where I aim to make people happy
* -Claim detection case: Yuriko Koike- *
[1st place] Candidate in the recent Tokyo gubernatorial election with the idea of ​​3.0 reform for the betterment of Tokyo
[2nd Place] Protecting the lives of Tokyo residents from all kinds of crises, including natural disasters
[3rd place] And we will make Tokyo the number one city in the world. * -Claim detection example: Mr. Shinji Ishimaru- *
[1st place] We have three pillars as our basic policy.
[2nd place] The first point is to make Tokyo government visible and understandable.
[3rd] Response to disaster risks

* ■Since 2016, the “questioning” tone has increased in political opinion broadcasts*

As a result of examining the political opinion broadcasts of all 58 candidates for the three years of 2016, 2020, and 2024, the number of “questioning” voices from candidates has increased. It was found that CLOSED questions, in which the listener can answer YES/NO, were used 1.54 times per person, an increase of 3.7 times compared to 2016. This trend can be seen in the 2022 House of Councilors election, and can be said to be a recent trend, as the number of speeches detected by political parties that are younger than their founding tends to increase.
The number of questions asked during political opinion broadcasts increases every year.

* ■The candidate with the highest number of votes is Shinji Ishimaru *

From the 2016 and 2020 data, the maximum score of 3,661,371 votes (Yuriko Koike in 2020) was given a perfect score, and talk features that increased in correlation with the number of votes were extracted. We have converted this characteristic into a mathematical formula and scored the political broadcasts of 2024 candidates.
Prediction of votes for 2024 Tokyo gubernatorial candidates
* – Shinji Ishimaru ranked No. 1 in terms of similarity to political poll broadcasts with high votes – *

Compared to past political opinion broadcasts of high vote-getters, this year’s political opinion broadcast had the closest
characteristics to Shinji Ishimaru. However, there is a difference in characteristics of 11%. The characteristics of Taro Yamamoto’s talk in 2020 are that there is a 7% deviation from Yuriko Koike in 2020.

* -Mr. Koike, different trends from last year-*

Yuriko Koike’s talk characteristics this year were significantly different from the last time, which received the most votes, with a 30% difference in overall characteristics. In particular, the number of main topics increased from one to two, and the dispersion of claims can be seen.
Given that political speeches that receive a high percentage of votes are characterized by a small number of claims (narrowed down), this year’s claims were rated as “weak.”

* -Weak arguments by Mr. Ninomiya, Mr. AI Mayer, Mr. Kusao, and Mr. Miwa-*

Mr. Ninomiya, Mr. AI Mayer, Mr. Kusao, and Mr. Miwa had lower vote-receiving characteristics than Mr. Oshikoshi in 2020, who received the lowest number of votes ever. What the four men had in common was that they included so many claims in a single political opinion broadcast that they accounted for more than 40% of the total amount of talk, and that they did not delve deeply into each claim. Furthermore, the person with the lowest percentage of claims this time was Mr. Tamogami at 9%.

* [Tokyo gubernatorial election 2024 survey summary captured through AI talk test] *
◯ Survey target: 3 gubernatorial election broadcasts totaling 86 talks so far (2024: 48 talks, 2020: 19 talks, 2016, 19 talks)
◯ Research method: “COG-PRESEN” and “COG-VIEW” tests using patented technology CogStructure
◯ Research organization: Cognity Co., Ltd.
◯ Publication site:
◯ Please include the following source information when publishing this research or processing and using graphs, etc.
  Cognity (2024) “Capturing the Tokyo gubernatorial election 2024 through AI talk test”
If possible, please contact us at See details

[Future of speech analysis for AI talk test]
This is the second time that we have announced the analysis of politicians’ speeches using the AI ​​Talk Test, after the 2022 House of Councilors election.
 2022 House of Councilors election/party claims analysis site:
For information on this Tokyo gubernatorial election, please see the posting site (
) will be updated after the election day, and additional survey results based on the vote count will be posted around August 1st, the day the new governor takes office.
In order to support each individual’s political participation through voting, we will continue to analyze the claims of each candidate and publish the survey results in order to provide “unbiased” information. *
*“Creating a society free of thinking bias through the power of technology” is Cognity’s mission, and we provide tools and services to encourage “bias-free” judgment in business situations, women’s empowerment, and regional revitalization. I am. *

* [About the AI ​​talk test “COG-PRESEN” and “COG-VIEW” used in this survey] *

There are two types of AI talk tests used in this survey:
“COG-PRESEN,” which detects the characteristics of individual political broadcasts, and “COG-VIEW,” which performs a comprehensive test over the past few years and allows trends and scoring. We are providing services to general companies. Using the only domestic and international technology (patented) that quantifies “quality” such as talk skills, it is possible to understand the skill level of an organization or team by comparing it with the talks of 60,000 people in the past.

In addition to the above two services, “COG-SERIES” also offers certification services related to human thinking and communication, including “COG-SALES,” “COG-COACH,” “COG-ESSAY,” and “COG-MR.” We use this information to help companies and managers solve organizational issues. Both services do not require the introduction of special equipment or systems, and can be implemented on the same day and at low cost.
AI test “Cog Series” that quantifies the quality of communication

Cognity Co., Ltd.
◯ Established: March 28, 2013
◯ Address: 1-2-208 Nishi-Oi, Shinagawa-ku, Tokyo 140-0015
◯ Web:
◯ Capital: 338,324,000 yen (including capital reserve)
◯ Employees: 109 people
◯ Representative: Rie Kono, Representative Director
◯ Business details
: Developing software to “remove cognitive biases”. Using a database of over 60,000 test results, we provide the “Cog Series,” an analysis service that models the trends of high performers and teams in customer service, presentations, meetings, etc., and quantitatively and comparatively evaluates them.
◇ Received the 22nd Telework Promotion Award Excellence Award ◇ Obtained ISMS certification (ISO27001) in June 2019, including remote workers ◇
The application was filed in 2015 and patented in August 2019 (Patent No. 6573321) for “Technology that calculates differences in the context and structure of talks and documents and predicts/recommends omissions, excesses and deficiencies”
◇ Since its founding in 2013, the company has adopted a “complete work-from-home system” and has been featured in many media outlets as a company that has successfully implemented the system.