baseline service quality

Baseline Service User Quality Floor

Document Positioning: This document explains the "quality floor" principle in Core Principle 2.6 and the application of the "human rights protection channel" in Core Principle 2.9 in scenarios of irreversible harm. Specific indicators and thresholds need to be operationalized according to domain and era conditions; this document only sets principle boundaries.

I. The Quality Floor Is Not a Sufficient Line for Justice

The quality floor should not be understood as: as long as baseline service users reach a certain bottom proportion, the system is already just.

A more accurate explanation is: the quality floor is a minimum alarm line, not a passing line; it is a trigger for mandatory rectification, not a reason for institutional self-satisfaction.

If a baseline service remains near the floor for the long term, it indicates that baseline service users have been in a disadvantaged position. Even if red collapse has not yet been triggered, quality catch-up, cause explanation, and feedback investment must be initiated.

II. Risk of Chronic Underconfiguration: The Floor Is Not a Long-Term Justice Line

The baseline service user quality floor must also address a more hidden failure: chronic underconfiguration.

Chronic underconfiguration is not the collapse of baseline service user services, nor the cancellation of baseline universal access. It refers to: baseline service users receiving safe, usable, compliant but lagging AI services over the long term, while high-risk threshold holders continuously receive stronger models, deeper explanations, faster feedback, higher freedom, and more complex toolchains. Baseline service users are not abandoned, but are steadily left behind.

This type of failure is easily obscured by floor indicators. As long as baseline service users do not fall below the minimum alarm line, the system may claim that "universal access still meets standards." But universal access cannot merely prevent people from falling; it must also prevent some people from never catching up.

Therefore, the quality floor must be bound to dynamic gap monitoring:

Cannot only compare accuracy rates, but must also compare explanation depth, update speed, tool freedom, personalization level, and public training opportunities.
Cannot only look at whether a particular year meets standards, but must also see whether the gap between baseline service users and high-risk threshold holders is continuously widening.
Cannot only look at whether services are available, but must also see whether baseline service users are gaining capability growth, governance participation, and capability development channels through services.
Cannot use "safety encapsulation" as a reason to long-term refuse delegating mature capabilities.

If baseline service users remain stable near the minimum line for multiple consecutive years, even if they have not fallen below the floor, a special review should be triggered. The focus of review is not "whether it is illegal," but "whether the system is compliantly manufacturing long-term backwardness."

III. Why a Number Is Still Needed

If numbers are completely rejected, quality commitments become empty words. The system can forever claim that "low-permission services are good enough" without providing verifiable standards.

The function of the quality floor is to turn "gentle abandonment" into a visible problem:

Are low-permission medical diagnoses significantly worse?
Does low-permission legal aid omit key steps?
Does low-permission educational AI use old models for the long term?
Are low-permission public services slower to respond, with fewer explanations and harder appeals?

Without verifiable standards, these gaps would be explained as individual cases, experiences, or resource shortages. With the floor, the system must at least acknowledge that gaps exist.

IV. Different Requirements for Three Types of Services

4.1 Irreversible Harm Scenarios: Applicable Human Rights Protection Channel

Including: emergency medical care, major disease diagnosis, criminal legal aid, personal safety warnings, disaster response, and critical public services.

In these scenarios, core service quality must not differ due to permission ladder position. Percentage quality floors do not apply to core judgments in such scenarios.

Core Service Equivalence Principle:

Diagnostic Conclusion: The final diagnostic judgment of medical AI must be equivalent, without accuracy differences due to patient permission ladder position.
Treatment Plan: Key treatment recommendations, medication plans, and surgery decisions must be equivalent.
Legal Strategy: Core defense strategies and key legal steps in criminal legal aid must be equivalent.
Safety Directive: Disaster evacuation directives and personal safety warning trigger conditions and content must be equivalent.

Allowed Differences Limited to Peripheral Expression:

Professional degree of explanatory language (terminology for doctors vs.通俗 explanations for patients).
Interface adaptation (simplified operation of basic interface vs. advanced functions of professional interface).
Richness of follow-up tracking services (not affecting initial rescue quality).

Hard Constraints on Peripheral Differences: Any peripheral differences must not affect the accuracy, completeness, and timeliness of core judgments. For example, simplified explanations cannot omit key risk information, and basic interfaces cannot delay the execution of key operations.

Resource Honesty: The human rights protection channel provides baseline protection, not unlimited resources. Public budget constraints may create upper limits for certain high-cost investments, but these limits must be public budget constraints, not permission ladder position discrimination. The system must publicly acknowledge this limit and continuously narrow the gap through feedback mechanisms. In any case, insufficient resources cannot be a legitimate reason for lowering core service standards—standards are non-negotiable, and insufficiency can only be addressed by expanding public investment.

4.2 Basic Development Scenarios

Including: general education, daily productivity tools, public information access, and basic administrative services.

These scenarios allow limited gaps, because high-risk threshold holders may involve more complex tools, higher freedom, and stronger customization capabilities. But gaps must meet three conditions:

Do not affect basic life opportunities.
Have clear capability development paths and compensation mechanisms.
Feedback funds prioritize narrowing gaps.

Rule: In basic development scenarios, the quality floor is a mandatory rectification line; long-term below higher standards requires public explanation and improvement plans.

4.3 High-Risk Operation Scenarios

Including: industrial control, financial model deployment, medical AI parameter modification, and critical infrastructure operation.

These scenarios do not apply the comparison of "low-permission service quality reaching a certain proportion of high permission." Because baseline service users should not have equivalent operational freedom.

What should be compared is: whether baseline service users can still obtain sufficient services, the right to know, the right to appeal, and capability development paths when they cannot operate high-risk systems.

Rule: In high-risk operation scenarios, the quality floor does not require equivalent control rights, but requires equivalent basic benefits and sufficient explanatory rights.

V. Relationship Between Quality Floor and Strong Guarantee

Rawlsian "maximization" requires all inequalities to most benefit the least advantaged. But in the field of AI control, "maximization" is difficult to operate: benefits cannot be single-quantified, and there are conflicts between safety, freedom, the right to know, and efficiency.

Stairway Universalism adopts "strong guarantee," not because it abandons bottom-line priority, but because it needs an enforceable trigger structure.

Strong guarantee contains at least four parts:

Quality Floor: Baseline service users cannot be systematically downgraded.
Right to Know: Baseline service users must know what they are restricted from and why.
Feedback Mechanism: The advantages of high-risk threshold holders must continuously flow back to the baseline universal layer.
Collapse Trigger: When quality gaps widen, the system must automatically review or reconstruct.

Therefore, the quality floor is only part of strong guarantee. Without the right to know, feedback, and collapse triggers, the floor degenerates into a technical performance indicator; with these mechanisms, it becomes part of the anti-abandonment structure.

VI. When Does It Constitute Institutional Failure

The following situations should be regarded as institutional failure even if they appear close to the floor:

Core services in irreversible scenarios differ due to permission ladder position. If in emergency medical care, criminal legal aid, and other scenarios, baseline service users are systematically given lower-standard core judgments (diagnostic conclusions, treatment plans, legal strategies, etc.), regardless of the gap size, it is regarded as absolute institutional failure. This failure cannot be defended by "cost constraints," "technical limitations," or "incremental improvement," and must be immediately rectified and held accountable.
Gaps concentrate in specific groups. If the elderly, disabled, low-income groups, or marginal regions are below average quality for the long term, overall quality达标 has no defensive force.
Baseline service users do not know gaps exist. Quality gaps without the right to know are implicit management, not universal access.
No catch-up mechanism. If the system only reports gaps without investing feedback funds and technological upgrades, the quality floor becomes formalism.
High-risk threshold holders continuously upgrade while baseline service users stagnate. Even if baseline service users maintain floor levels in the short term, if the trend continues to decline, a review should be triggered in advance.
Long-term stable underconfiguration. If baseline service user services are safe, usable, and compliant for the long term, but always lag behind the mature capabilities of high-risk threshold holders, and lack clear delegation timelines, it should also be regarded as chronic institutional failure.

VII. Human Rights Protection Channel Mechanism

7.1 Design Principles

The human rights protection channel is a pre-institutional arrangement of Stairway Universalism, not an internal ladder position design of the stairway system. It addresses "under what circumstances must the capability-permission matching principle yield to non-derogable human rights."

Core judgment: If a hierarchical system systematically places people in life-threatening situations, it lacks legitimacy. This is not a negation of the stairway system, but a demarcation of its applicable boundaries.

7.2 Trigger Conditions

The human rights protection channel automatically activates, without individual application or risk threshold certification. Trigger conditions include:

Emergency Medical Care: Acute diseases, trauma, life-threatening vital signs, and other scenarios requiring immediate intervention.
Major Disease Diagnosis: Diagnosis processes for cancer, cardiovascular and cerebrovascular diseases, and other diseases that may directly threaten life.
Personal Safety Warning: Violence threats, natural disaster warnings, public safety alarms, etc.
Disaster Response: Evacuation, rescue, and medical resource allocation in large-scale disasters such as earthquakes, floods, and epidemics.
Criminal Legal Aid: Defense services when defendants face deprivation of personal freedom or major penalties.

Not Trigger Conditions (personal choices, do not enter the channel):

Elective surgery, cosmetic medicine, routine physical examinations.
Commercial legal consultation, civil disputes (not related to personal freedom).
Education, career development, financial planning, and other non-survival services.

7.3 Service Provision Methods

Bypass Conventional Certification: Anyone automatically accesses the highest standard core AI assistance, without going through conventional risk threshold certification processes.
Unified Core Standard: Key outputs such as diagnostic conclusions, treatment plans, legal strategies, and safety directives must be equivalent to those of high-risk threshold holders.
Differentiated Expression: Explanation forms and interface adaptation can be differentiated, but must not affect core judgments. For example, basic interfaces can simplify operation steps, but cannot simplify the accuracy of medical judgments.
Public Fund Bearing: Costs are borne by public funds, not traced back to individual permission ladder positions. This ensures that economic capacity does not affect emergency rescue quality.

7.4 Responsibility Attribution

AI-assisted decision-making responsibility in emergency scenarios is borne by public institutions, not individuals. This means if medical AI gives a wrong diagnosis in emergency care, the responsible subject is the public medical system providing that AI, not the patient.
Additional obligations of high-risk threshold holders: Even if emergency patients obtain equivalent AI diagnosis through the human rights protection channel, high-permission medical professionals still bear supervision and review obligations. The channel provides tool equality, not professional responsibility exemption.

7.5 Anti-Abuse Mechanisms

Scenario Audit: Medical institutions and public systems must establish scenario audit mechanisms to ensure that the human rights protection channel is only used for true irreversible harm scenarios. Abuse (such as disguising elective surgery as emergency care) will trigger audit and accountability.
Data Logging: All services through the human rights protection channel must be logged, auditable, and traceable to prevent systematic abuse.
Public Supervision: Human rights protection channel usage data (trigger times, service types, resource consumption) must be regularly publicized and subject to public supervision.

VIII. Summary

The significance of the quality floor is not to claim that "a certain proportional gap is just," but to refuse to let gaps remain invisible.

Stairway Universalism's strong guarantee should not be understood as a number, but as a set of mandatory mechanisms: discovering gaps, explaining gaps, narrowing gaps, compensating gaps, and triggering reconstruction when gaps are unacceptable.

Baseline service users not falling below the floor does not mean the system is already just. Long-term stable underconfiguration itself may also constitute failure.

Institutional Engineering Honesty: The quality floor is not the end of justice, but the starting point where the system can no longer pretend there is no problem.