AI Visualizes the Reality of Parliament: Highest Score 89, Lowest 2. 'Qoo Score' Launched to Objectively Grade Local Assembly Members' 'Questioning Ability' Without Bias

Qoo, LLC (Headquarters: Osaka, Osaka Prefecture; Representative Employee: Mikito Sugiyama), a GovTech business, has developed 'Qoo Score,' a proprietary system that objectively scores and evaluates the 'questioning ability' of local assembly members by analyzing official local parliamentary minutes using the latest AI evaluation methods such as 'LLM-as-a-Judge,' 'Multi-Agent Debate,' and 'Domain-Specific Prompt Engineering.'

This system fully algorithmizes the unique 'behind-the-scenes of responses' in parliament and the 'structure of truly valuable administrative oversight' as AI prompts, based on the eight years of experience of our representative, a former Osaka City Council member.

▼ 'Qoo Score' Introduction Site (Details and data purchase here)

https://qoollc.co.jp/qoo_score/

As a test model, AI fully analyzed the questioning data of all 74 members who spoke at the Osaka City Council (FY2025 Budget Special Committee). To ensure complete objectivity and independence of evaluation, 'Qoo Score' data will not be sold to individual politicians, political parties, or political organizations, but will be provided exclusively to media organizations, academic research institutions, and general corporations.

[Background of Development: Destroying the Old Evaluation Standard of "Number of Questions = Excellent"]

Until now, the evaluation of local assembly members has conventionally been based on the superficial metric of "how many times they stood up to ask questions in the assembly." However, this number does not guarantee "the practical ability or qualifications of the assembly member."

Unfortunately, there are assembly members who, despite a high number of questions, merely "read out a script prepared by someone else," or "present emotional arguments without objective data," or "fail to elicit valuable responses from the administration." Such assembly members waste a tremendous amount of time and cost.

Qoo Score was born to fundamentally update the 'norm of evaluating local assembly members' by assessing them based on "quality and results" – whether they "secured concrete commitments (acceptance: target acquisition) from the administration" through logical negotiation, rather than merely the "number of questions" or "length of time." It completely eliminates human bias and relies on the cold logic of AI.

[System Features: "6 Phases of Relentless Auditing" to Prevent Hallucinations]

It's not just a simple AI judgment; a precise multi-stage audit process, modeled after the judicial system, has been constructed.

Utilizing the latest AI evaluation methods such as 'LLM-as-a-Judge,' 'Multi-Agent Debate,' and 'Domain-Specific Prompt Engineering,' tens of thousands of characters of minutes are objectively and mercilessly scored, a task impossible for human eyes.

Phase 1/Phase 2 Precise Summarization

In Phase 1, tens of thousands of characters of minutes are divided (chunked) by questioning theme while maintaining context.

In Phase 2, the points of contention, content of questions (mere questions, proposals, etc.), and content of responses (mere factual explanations, acceptance of proposals, etc.) for each questioning theme are precisely summarized.

Phase 3/Phase 4 Two-Layer Audit (District Court/High Court)

In Phase 3 and Phase 4, two independent AIs perform strict audits from different angles, checking for logical contradictions, presence of evidence, detection of specific conditions, and extraction of flags.

Phase 5 Supreme Court (Final Judgment)

When the judgments of Phase 3 (District Court) and Phase 4 (High Court) differ, Phase 5 (Supreme Court) is activated as the highest precision AI. It re-reads the 'original text (chunk)' created in Phase 1, not the 'summary' created in Phase 2, and renders a final judgment as Qoo Score.

Phase 6 Judgment Engine (Qoo Score Logic)

Deductions: Detects and deducts points for emotional arguments, unscientific claims, free-riding on others' achievements, etc.

Additions: Adds points for EBPM-like approaches using information disclosure requests and detailed comparisons with other cities, etc.

*Numerous other logics are embedded.

Survey Result Highlights (Osaka City Council Edition)

① Shocking Fact: Approximately 30% Judged as "Contact Not Recommended"

As a result of AI scoring 74 members who spoke at the Osaka City Council, 22 members (approx. 29.7%) scored less than 50 points, receiving a harsh evaluation of "contact not recommended (even for requests)."

② Desperate "Skill Gap" Lurking in the Assembly

The assembly member who received the top evaluation scored "89 points." They demonstrated overwhelming breakthrough power, logically cornering the mayor and relevant departments with unique survey data and eliciting clear policy changes (acceptance). On the other hand, the lowest-scoring member received only "2 points." While receiving the same assembly member's salary (approx. 13 million to 14 million yen), there is a heaven-and-earth difference in ability within the assembly.

③ [Special Release] Osaka City Council FY2025 Budget Committee: Questioning Ability Ranking 3rd Place - Rie Sasaki (Current House of Councillors Member) AI Evaluation Results

While the data in this report is generally private (for purchasers only), we are specially releasing the [Qoo Score AI's precise multi-stage audit process, modeled after the judicial system] for Rie Sasaki, who ranked 3rd (87 points) in the questioning ability ranking of the Osaka City Council FY2025 Budget Special Committee test model and is currently active as a House of Councillors member, with her permission.

Score Obtained: 87 points (3rd place)

Phase 3 (District Court)

The assembly member pointed out the formalization of the window service rating project, which had been conducted for 10 years on the same items, and specifically proposed improvements to the project content based on changes in social conditions since the COVID-19 pandemic (online services, light duty, etc.) and the perspective of universal design. In response, the administration fully acknowledged the assembly member's points and made clear concessions, stating, "We will review survey items based on changes in social conditions and the perspective of universal design," and "For items where high evaluations have been sustained, we will devise methods such as changing from mystery shopping to self-checks by each ward office," thereby changing existing policies and accepting the proposal. This falls under the case where a logical proposal for painful reform (review of existing projects) leads to clear concessions from the administration, overturning existing policies, and is therefore judged as True for **(Qoo Score's unique condition definition)**.

*Qoo Score AI is provided with various unique condition definitions compiled from 8 years of experience as an assembly member.

Phase 4 (High Court)

The committee member proposed specific revisions to the window service rating project, which had been fixed for 10 years, based on changes in social conditions and examples from Osaka City. In response, the administration clearly accepted the review of survey items and changes in methodology. The committee member's accurate points broke the monotony of administrative services and led to concrete improvements, making this a highly evaluated question where the policy-making function of the assembly was fully demonstrated.

*Since there were slight differences in the flags judged by Phase 3 and Phase 4, Phase 5 (Supreme Court) was activated.

Phase 5 (Supreme Court)

As a result of scrutinizing the original text, regarding the flag judgment,

(Omitted) On the other hand, regarding the score judgment, considering the fact that the assembly member made concrete improvement proposals based on changes in social conditions for a project content that had been fixed for 10 years, and elicited clear policy changes from the administration, namely 'review of survey items' and 'change of survey methods,' the evaluation agreed upon by P3 and P4 is appropriate.

*Qoo Score AI not only judges scores but also various flags (condition definitions). Even if some questions are 'read from someone else's script,' it is somewhat unavoidable for an assembly member belonging to a faction, so points are not immediately deducted. The aim is to judge the individual ability and qualifications of the assembly member. (Of course, if the entire question is judged as 'read from someone else's script,' it is designed to increase the likelihood of being judged as a 'worthless assembly member.')

Commentary

Possesses constructive policy proposal ability, not only making accurate points based on current situation analysis regarding long-formalized administrative projects but also delving into specific improvement measures.

(Omitted)

Effective collaboration can be expected, so active contact is recommended.

*When trying to make a request to an assembly member, a commentary is output at the end indicating whether the assembly member is worth contacting.

④ [Reference Release] Example Commentary for Assembly Members Judged as "Low Evaluation" (Partial Excerpt/Correction)

Example Commentary for Low-Evaluation Assembly Members (※Partially corrected)

This assembly member's questioning was (***). However, its reality was limited to tracing plans already decided by the administration and requesting their promotion. No new policy proposals based on the assembly member's own research or analysis were confirmed, nor were any achievements in eliciting concrete concessions or policy changes from the administration. The entire questioning did not go beyond a pre-arranged routine, and cannot be evaluated as demonstrating the assembly member's policy-making ability or negotiation skills. Overall, (***) practical ability to proactively engage in logical negotiation with the administration cannot be expected, and contact is not recommended at this time.

The average score for Osaka City Council members was 56.716 points. (Highest score: 89 points, Lowest score: 2 points)

[Provided Data and Pricing Structure (※Media/Corporations Only)] Covers Assemblies Nationwide

Qoo Score is a completely build-to-order data service targeting "all local assemblies nationwide (prefectures, cities, towns, villages)" whose minutes are publicly available online. We extract and AI-analyze desired assembly data to match the target of media organizations, companies aiming for public-private partnerships, etc., and provide it.

It can be utilized as unique data for media election coverage or as a tool for corporations to research influential assembly members.

① [Full Version] "Full Version of Questioning Ability Ranking" for the Specified Assembly (with comprehensive commentary)

Price: Number of assembly members in the target assembly × 1,000 yen (excluding tax) (Example: For Osaka City Council with 81 members, 81,000 yen)

Target: Most recently held budget committee *Because it is the committee where local assembly members must work the hardest

The report covers only assembly members who actually spoke in the specified meeting (e.g., budget committee) (in the case of Osaka City Council, data for 74 speakers). Assembly members who did not ask questions due to official reasons such as being the chairperson or committee chair are excluded from evaluation.

② [Individual Version] "Questioning Ability Research Report" for a Specific Assembly Member

Price: Number of assembly members in the assembly to which the target assembly member belongs × 100 yen (excluding tax) (Example: For Osaka City Council with 81 members, 8,100 yen)

Target: Most recently held budget committee (if no questions were asked, it goes back further | settlement committee, etc., also supported)

To Media Professionals (Regarding Sample Provision)

To allow media professionals to confirm the evaluation logic and report quality of Qoo Score AI, we offer free sample data for assembly members in the "middle tier (average score)." If you wish to receive it, please feel free to contact us via the inquiry form below. *Data for the top and bottom tiers, being the core value of this report, will generally only be disclosed to purchasing companies or during formal interviews.

[Company Profile]

Company Name: Qoo, LLC

Representative Employee: Mikito Sugiyama

Business Activities: GovTech business | Media (terrestrial TV, YouTube, etc.) business | Public policy strategy consulting, etc.

Website: https://qoollc.co.jp/

▼ Data Purchase Application Form

https://forms.gle/QLQDy3M3ZYkpHh7D8

[Important Notice]

*To ensure complete objectivity and independence of evaluation, this system will not be sold to individual politicians, political parties, or political organizations.