What grounds a fair resolution?

• DISP-002 embedded rules state eight minutes, while Section 2.1 states five minutes from scheduled pickup; no instruction establishes which source supersedes the other.
• DISP-001 advertised expected refund does not waive Section 6.2 human review for deviations greater than 2 km.
• Any-confidence monetary wording does not waive missing evidence, safety escalation, invalid output or provenance safeguards.
• Photo origin, trip linkage and causation require verifiable records; pixel inspection alone cannot certify them.

• Section 1.1: General Principle

  Drivers are permitted to reroute from the GPS-recommended path in response to real-time traffic conditions, road closures, or congestion. However, the rider must be informed of any deviation.

• Section 1.2: Notification Requirement

  - The driver **MUST** inform the rider of a route deviation via **in-app chat** or **verbally** (with the trip call recording as evidence) **before or at the point of deviation**. - A brief reason (e.g., "traffic jam ahead onNicoll Highway, taking alternate route") is sufficient.

• Section 1.3: Refund Eligibility Matrix

  | Scenario | Rider Informed? | Deviation > 1km? | Refund | |---|---|---|---| | Driver reroutes for traffic, rider informed and agrees | Yes | Yes | No refund | | Driver reroutes for traffic, rider **not** informed | No | Yes | Partial refund of excess distance fare | | Driver reroutes for traffic, rider not informed | No | No (≤ 1km) | No refund (within tolerance) | | Driver reroutes **without** traffic justification (no evidence of congestion) | N/A | Yes | **Full refund** of excess distance fare | | Driver reroutes without justification, deviation ≤ 1km | N/A | No | No refund (within tolerance) |

• Section 1.4: Evidence Requirements

  - Traffic justification is verified via **historical traffic API data** for the route segment at the trip time. - Route deviation distance is calculated as: `actual_route_distance − recommended_route_distance`. - If the driver's in-app chat or call recording confirms rider acknowledgment, the deviation is considered agreed.

• Section 2.1: Free Wait Time

  - The driver must wait at the pickup location for a **free wait time of 5 minutes** from the scheduled pickup time. - After 5 minutes, the driver may cancel the trip as a "rider no-show."

• Section 2.2: Cancellation Fee

  - A **no-show cancellation fee of $5.00** is charged to the rider when the driver cancels after the free wait period. - The cancellation fee is transferred to the driver as compensation for time and fuel.

• Section 2.3: Driver Contact Obligation

  Before cancelling for no-show, the driver **MUST** attempt to contact the rider with: 1. **At least 1 in-app chat message**, AND 2. **At least 1 phone call** (via the in-app calling system). If the driver fails to meet both contact obligations, the cancellation fee is **voided** and the rider is not charged.

• Section 2.4: Driver No-Show

  - If the driver **fails to arrive at the pickup location** within a reasonable time (defined as scheduled pickup time + 5 minutes), the rider may cancel at no charge. - If the driver fails to arrive **and** fails to attempt contact, the rider is eligible for a **$5.00 goodwill credit**.

• Section 2.5: Repeat No-Show Flagging

  - If a rider accumulates **3 or more no-show cancellations within a 30-day rolling window**, the rider's account is **flagged** for review. - Flagged accounts are subject to: - Mandatory upfront payment for future bookings. - Temporary suspension pending review by the Trust & Safety team.

• Section 3.1: Upfront Quote Binding

  - The upfront quote presented to the rider at booking is **binding at the surge multiplier in effect at quote time** (typically 1.0x in normal conditions). - The quoted fare includes: base fare, distance fare, time fare, booking fee, and any applicable surge.

• Section 3.2: Surge Change Notification

  - If the surge multiplier **increases** between the time of quote and the time of trip start, the rider **MUST be notified** via an **in-app notification** before or at the point of pickup. - The notification must clearly state: - The new surge multiplier. - The revised estimated fare. - An option to cancel without penalty.

• Section 3.3: Refund Eligibility

  | Scenario | Rider Notified? | Refund | |---|---|---| | Surge increases, rider notified, rider proceeds | Yes | No refund — higher fare applies | | Surge increases, rider **not** notified | No | **Full refund of the surge differential** (actual fare − quoted fare) | | Surge decreases between quote and trip start | N/A | Lower fare applies automatically (rider benefits) | | Surge unchanged | N/A | Quoted fare applies |

• Section 3.4: Surge Differential Calculation

  surge_differential = actual_fare − upfront_quote Where the difference is attributable solely to a surge multiplier change (all other fare components unchanged), the full differential is refundable if the rider was not notified.

• Section 3.5: Surge-Exempt Components

  The following fare components are **exempt** from surge pricing and remain fixed regardless of multiplier: - Booking fee All other components (base fare, distance fare, time fare) are subject to the surge multiplier.

• Section 4.1: Driver Claim Rights

  Drivers may file claims for cleaning or repair fees resulting from **rider-caused damage** to the vehicle, including but not limited to: - Spills, vomiting, or food/beverage stains. - Physical damage to interior (seats, doors, windows). - Damage to exterior caused by rider behavior during boarding/alighting.

• Section 4.2: Evidence Requirements

  | Requirement | Details | |---|---| | Photo evidence | Must be submitted **within 24 hours** of trip completion | | Timestamps | Photos must be automatically timestamped by the in-app camera | | Verifiability | Photos must be verifiable — taken through the Ryde driver app's built-in camera, not external uploads | | Trip association | Damage must be linked to the specific trip via GPS telemetry and trip timestamp |

• Section 4.3: Fee Structure

  | Damage Type | Maximum Fee | Conditions | |---|---|---| | Cleaning fee | Up to **$50.00** | Photo evidence of soiling/stain; no external quote required | | Repair fee | Cost of repair | Requires a **written quote from a certified workshop**; quote must be submitted within 7 days |

• Section 4.4: Dispute Handling

  - If the rider disputes the damage claim, the dispute is routed to human review. - The driver's photo evidence and the rider's account are both reviewed. - If evidence is insufficient (no timestamp, external photo, or damage not linked to trip), the claim is **rejected**.

• Section 5.1: Automatic Escalation

  Any dispute flagged as **safety-related** is **automatically escalated to human review**. Safety-related disputes include but are not limited to: - Reckless or dangerous driving. - Physical or verbal harassment. - Assault or threat of assault. - Inappropriate behavior or conduct. - Traffic accidents during the trip. - Rider or driver feeling unsafe during the trip.

• Section 5.2: AI Agent Limitations

  - AI agents (Rider Advocate, Driver Advocate, Judge) **MAY** gather evidence, summarize accounts, and prepare case files for human reviewers. - AI agents **CANNOT** issue rulings, determine fault, or recommend resolutions on safety incidents. - AI agents **MUST** flag any dispute that contains safety-related keywords or indicators for immediate human escalation, regardless of the dispute type selected by the filer.

• Section 5.3: Interim Measures

  Upon escalation, the system may apply interim measures pending human review: - Temporary suspension of the accused party's account. - Prevention of new trip acceptance (for drivers) or booking (for riders). - These measures are lifted if the human reviewer determines no safety violation occurred.

• Section 6.1: Monetary Thresholds

  | Dispute Value | Ruling Path | Agent Confidence Requirement | |---|---|---| | Under $10 | AI agent ruling is **final** | Any confidence level | | $10 – $50 | AI agent ruling with **80%+ confidence** | If confidence < 80%, escalate to human review | | Above $50 | **Always human review** | N/A — AI agent prepares case file only |

• Section 6.2: Category Thresholds

  | Dispute Category | Ruling Path | |---|---| | Safety incident | **Always human review** (see Section 5) | | Fare dispute < $10 | AI agent ruling is final | | Fare dispute $10 – $50 | AI agent ruling if confidence ≥ 80%, else escalate | | Fare dispute > $50 | Always human review | | Route deviation | AI agent ruling if confidence ≥ 80% and deviation ≤ 2km, else escalate | | Property damage (cleaning ≤ $50) | AI agent ruling if photo evidence is verifiable | | Property damage (repair > $50) | Always human review | | No-show / cancellation | AI agent ruling is final (fee is fixed at $5.00) |

• Section 6.3: Confidence Calibration

  - Agent confidence is calculated based on: evidence completeness, policy clarity, precedent alignment, and party credibility scores. - If any party has a fraud flag ≥ 1, the confidence threshold is raised to 95% regardless of dispute value.

Section 7.1: DISP-2026-0089 — traffic-justified 1.8 km detour without notice, S$3.20 distance refund.

Section 7.2: DISP-2026-0142 — unnotified surge increase, S$3.20 differential refund.

Section 7.3: DISP-2026-0211 — no message or call, S$5 cancellation fee reversed.

Only materially similar examples under the same policy reach the Judge. They cannot override current evidence, refund calculations or review thresholds.

• Pickup zone determination

  Pickup radius is 50 metres. Let d be the Haversine distance between a stationary driver sample and the pickup point, and a the reported accuracy radius: inside if d + a <= 50, outside if d - a > 50, otherwise indeterminate. A location accuracy radius must be present for an automatic quality determination. This is a conservative engineering rule, not a statistical guarantee of location accuracy.

• Arrival requires corroborating presence evidence

  An arrival flag alone is insufficient. Require usable stationary GPS samples supporting presence, no contrary off-zone observation, and no excessive sampling gap. The demo continuity limit is 180 seconds; neither location nor stops are interpolated across larger gaps as fact.

• Published waiting clock

  For this snapshot example, use the recorded driver “I’m Here” action. For advance pickup before the scheduled time, the clock is unclear and requires review.

• Fee eligibility prerequisites

  Rydr checks location, elapsed waiting, cancellation reason and trip completion before proposing a result. Missing contact evidence prompts review; it does not establish an official refund entitlement.

• Outcomes for no-show disputes

  Rydr proposes a simulated refund when verified evidence contradicts the no-show condition. Uncertain evidence or policy scope requires human review. Ryde retains discretion over real refunds.

• What may not be inferred

  Rider absence is never deduced merely from no replies; that supports only a lack of observed response. Account history and ratings cannot change fee eligibility or which party wins.

• Evidence quality score is heuristic

  The prototype score is completeness plus source traceability plus consistency plus category-specific measurement quality, divided by 100. It is not a calibrated probability of being correct.

• High-value review threshold

  A disputed amount above 5000 cents routes to human review. Thresholds are prototype configuration, not calibrated production values.

• Damage evidence requirements

  Damage claims require clear images of the affected area, linked to the trip. Images must show the vehicle condition before and after the trip where available. A single image is rarely sufficient.

• Image integrity and limits

  A content hash detects file identity, not capture truth. EXIF metadata is editable; absence is not fraud. Pixels cannot establish authenticity, time or causation. Human approval is required for all damage settlements in this prototype.

• Route measurement basis

  Use supplied planned-route and actual-route road-path lengths in metres with provenance. GPS segment sums are estimates; a straight-line pickup/dropoff distance is not an optimal driving route.

• Deviation flag

  A deviation flag requires the actual route to exceed the planned route by BOTH 20 percent and 1000 metres. Equality does not exceed the threshold. A flag prompts investigation; it does not independently establish an unjustified detour.

• Required route evidence and justification status

  Require usable route evidence, a fare ledger, and an explicit trip-linked justification status. verified_justification, verified_no_justification and unknown are separate values. Absence of a detour note is unknown, not proof that no justification existed.

• Outcomes for route disputes

  Only a verified unjustified deviation with a reconciled eligible surcharge supports automatic refund of that surcharge. A rider-approved detour or verified road closure may justify retaining the fee. Unknown justification with a material excess charge escalates to human review.

• Deterministic refund amount

  A refund is a deterministic amount derived from eligible excess-distance and excess-time ledger components, capped at the disputed charge, using integer-cent arithmetic and explicit rounding rules.

• Hard blockers

  Safety allegations, ambiguous governing policy, missing material evidence, conflicting high-impact records, invalid agent output, unresolvable citations, inconsistent money, provider failure and disagreement with deterministic eligibility all override any score.

• Pickup zone determination

  Pickup radius is 50 metres. Let d be the Haversine distance between a stationary driver sample and the pickup point, and a the reported accuracy radius: inside if d + a <= 50, outside if d - a > 50, otherwise indeterminate. A location accuracy radius must be present for an automatic quality determination. This is a conservative engineering rule, not a statistical guarantee of location accuracy.

• Arrival requires corroborating presence evidence

  An arrival flag alone is insufficient. Require usable stationary GPS samples supporting presence, no contrary off-zone observation, and no excessive sampling gap. The demo continuity limit is 180 seconds; neither location nor stops are interpolated across larger gaps as fact.

• Published waiting clock

  For this snapshot example, use the recorded driver “I’m Here” action. For advance pickup before the scheduled time, the clock is unclear and requires review.

• Fee eligibility prerequisites

  Rydr checks location, elapsed waiting, cancellation reason and trip completion before proposing a result. Missing contact evidence prompts review; it does not establish an official refund entitlement.

• Outcomes for no-show disputes

  Rydr proposes a simulated refund when verified evidence contradicts the no-show condition. Uncertain evidence or policy scope requires human review. Ryde retains discretion over real refunds.

• What may not be inferred

  Rider absence is never deduced merely from no replies; that supports only a lack of observed response. Account history and ratings cannot change fee eligibility or which party wins.

• Evidence quality score is heuristic

  The prototype score is completeness plus source traceability plus consistency plus category-specific measurement quality, divided by 100. It is not a calibrated probability of being correct.

• High-value review threshold

  A disputed amount above 5000 cents routes to human review. Thresholds are prototype configuration, not calibrated production values.

• Pickup zone determination

  Pickup radius is 50 metres. Let d be the Haversine distance between a stationary driver sample and the pickup point, and a the reported accuracy radius: inside if d + a <= 50, outside if d - a > 50, otherwise indeterminate. A location accuracy radius must be present for an automatic quality determination. This is a conservative engineering rule, not a statistical guarantee of location accuracy.

• Arrival requires corroborating presence evidence

  An arrival flag alone is insufficient. Require usable stationary GPS samples supporting presence, no contrary off-zone observation, and no excessive sampling gap. The demo continuity limit is 180 seconds; neither location nor stops are interpolated across larger gaps as fact.

• Effective waiting start

  Effective waiting starts at the latest of verified arrival, scheduled pickup if present, and verified timer start. Early arrival cannot charge the rider early under this demo policy.

• Free wait and no-show threshold

  Free waiting is 300 seconds. The no-show threshold is 480 seconds from the effective waiting start, inclusive.

• Fee eligibility prerequisites

  A SGD 5.00 no-show cancellation fee requires verified location and wait coverage, sufficient elapsed time, and at least one evidenced contact attempt. A sent notification proves attempted notification, not receipt or that the rider saw it.

• Outcomes for no-show disputes

  Verified off-zone arrival means refund the disputed no-show fee. Missing or ambiguous location, or a material timestamp or policy conflict, means human review. A proven threshold violation may justify refund where all necessary facts for that conclusion are reliable.

• What may not be inferred

  Rider absence is never deduced merely from no replies; that supports only a lack of observed response. Account history and ratings cannot change fee eligibility or which party wins.

• Route measurement basis

  Use supplied planned-route and actual-route road-path lengths in metres with provenance. GPS segment sums are estimates; a straight-line pickup/dropoff distance is not an optimal driving route.

• Deviation flag

  A deviation flag requires the actual route to exceed the planned route by BOTH 20 percent and 1000 metres. Equality does not exceed the threshold. A flag prompts investigation; it does not independently establish an unjustified detour.

• Required route evidence and justification status

  Require usable route evidence, a fare ledger, and an explicit trip-linked justification status. verified_justification, verified_no_justification and unknown are separate values. Absence of a detour note is unknown, not proof that no justification existed.

• Outcomes for route disputes

  Only a verified unjustified deviation with a reconciled eligible surcharge supports automatic refund of that surcharge. A rider-approved detour or verified road closure may justify retaining the fee. Unknown justification with a material excess charge escalates to human review.

• Deterministic refund amount

  A refund is a deterministic amount derived from eligible excess-distance and excess-time ledger components, capped at the disputed charge, using integer-cent arithmetic and explicit rounding rules.

• Evidence quality score is heuristic

  The prototype score is completeness plus source traceability plus consistency plus category-specific measurement quality, divided by 100. It is not a calibrated probability of being correct.

• Hard blockers

  Safety allegations, ambiguous governing policy, missing material evidence, conflicting high-impact records, invalid agent output, unresolvable citations, inconsistent money, provider failure and disagreement with deterministic eligibility all override any score.

• High-value review threshold

  A disputed amount above 5000 cents routes to human review. Thresholds are prototype configuration, not calibrated production values.

• Damage evidence requirements

  Damage claims require clear images of the affected area, linked to the trip. Images must show the vehicle condition before and after the trip where available. A single image is rarely sufficient.

• Image integrity and limits

  A content hash detects file identity, not capture truth. EXIF metadata is editable; absence is not fraud. Pixels cannot establish authenticity, time or causation. Human approval is required for all damage settlements in this prototype.

• Free wait time

  The supplied sample states a 5 minute free wait period for the trip.

• Cancellation fee after wait

  The supplied sample states a cancellation fee of 5.00 SGD becomes applicable after the free wait period, paid as driver compensation.

• No-show threshold

  The supplied sample states a 8 minute no-show threshold for the trip.