Enterprise AI Analysis
Communicating Uncertainty in Arrival Time Predictions for Public Transport: A Comparison of Point and Interval Forecasts
Authors: Alice Rollwagen, Alexander Horn, Stefanie Schmidtner, Andreas Riener
In public transport, arrival times are typically communicated as point forecasts, aiming for precision. However, current prediction models often fail to provide reliable estimates due to unpredictable events, leading to passenger dissatisfaction from perceived inaccuracies. This study investigates interval forecasts as an alternative to better communicate uncertainty in arrival times. Findings indicate that interval forecasts improve the communication of uncertainty and that user satisfaction is primarily driven by waiting time, moderated by the forecast concept. Point forecasts were only well received when the bus arrived as predicted; otherwise, users preferred broader interval forecasts, valuing accuracy over precision.
Executive Impact & Strategic Imperatives
Misleadingly precise arrival time predictions in public transport erode passenger trust and satisfaction. This research provides a crucial pathway for enterprise-level improvements by advocating for transparent uncertainty communication through interval forecasts, directly addressing a core user experience challenge.
0
Increase in Satisfaction for Late Arrivals with Interval Forecasts
0
Approx. Interval Width for 95% Confidence (15 min forecast)
0
Reduced Perceived Waiting Time with Real-time PIS
0
Point Forecast Pragmatic Quality Score (vs 0.014 for Interval)
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
The Challenge of Prediction Uncertainty in Public Transport
Modern real-time passenger information systems (PIS) significantly improve public transport (PT) experience, yet they often rely on point forecasts (e.g., "arrival in 7 min") which inherently lack the ability to communicate uncertainty. Unpredictable events (traffic, weather, passenger counts) lead to deviations from these precise predictions, resulting in passenger dissatisfaction. Research indicates that users prefer transparent communication of uncertainty over misleading precision.
For instance, sophisticated models like ConvLSTM achieve an RMSE of 2.66 minutes at a 15-minute forecast horizon. Assuming normal error distribution, this translates to a 95% confidence interval of approximately 10 minutes wide. This gap between desired precision and real-world accuracy highlights the need for a better communication strategy.
Online Study Design and Data Simulation
An online questionnaire was conducted with 69 respondents (aged 18-54) to compare the interpretation and user experience of point versus interval forecasts. Two PIS mock-ups were created: a point forecast mirroring existing German bus-stop displays, and an interval forecast simulated using real historical data from Ingolstadt, Germany (Jan-Dec 2022). The interval forecast converted point predictions into a range based on the overall distribution of delays, with the width converging to zero as the bus approached the stop.
Participants evaluated scenarios involving early, exact, and late bus arrivals for two different bus lines (Line 40 with narrower time spans, Line 16 with wider time spans) and rated their satisfaction. User experience was assessed using the standardized UEQ-S questionnaire, measuring pragmatic and hedonic quality.
User Perception, Satisfaction, and Experience Analysis
The study found that interval forecasts significantly improved the communication of uncertainty; participants formed clearer expectations of arrival times with ranges. While point forecasts yielded higher satisfaction for exact arrivals, interval forecasts led to significantly higher satisfaction when buses arrived early or late, demonstrating that users value accuracy over misleading precision. For example, for a 18-minute late arrival on Line 16, point forecast satisfaction was 1.90 while interval forecast satisfaction was 2.93. This indicates that providing a range mitigates negative impacts of deviations.
However, pragmatic quality scores from the UEQ-S showed point forecasts rated better (1.996 vs. 0.014 for interval forecasts), suggesting that the directness of a point forecast is still perceived as more efficient and perspicuous, despite its real-world shortcomings when inaccurate. This highlights a paradox: users prefer precise information even when it leads to frustration.
Strategic Implications for PIS and Future Research
This research strongly suggests that public transport operators should consider implementing interval forecasts to provide more realistic arrival time expectations. While point forecasts are often preferred for their perceived precision, their failure to reflect real-world variability can lead to significant passenger dissatisfaction. Interval forecasts, by explicitly communicating uncertainty, align better with actual system capabilities and passenger needs for accuracy.
Future work could explore alternative visualizations for uncertainty (e.g., "5 min ± 2 min" or displaying the mode within the interval), investigate varying interval widths and forecast horizons, and conduct A/B tests with real passengers at bus stops using digital displays. It also emphasizes the need to balance the cognitive preference for simplicity with the critical need for truthful, accurate information.
54%
Higher User Satisfaction for Late Arrivals with Interval Forecasts (Line 16, 18 min)
| Feature | Point Forecasts (Traditional) | Interval Forecasts (Proposed) |
|---|---|---|
| Communication of Uncertainty |
|
|
| User Satisfaction (Exact Arrival) |
|
|
| User Satisfaction (Early/Late Arrival) |
|
|
| Perceived Cognitive Load |
|
|
| Alignment with Real-world Variability |
|
|
Passenger Perception & Expectation Flow
Forecast Displayed (Point or Interval)
→
Passenger Forms Arrival Expectation
→
Bus Arrives (Actual Time)
→
Comparison to Expectation
→
Experience Satisfaction/Dissatisfaction
Case Study: Enhancing Public Transport Experience with Predictive AI
Scenario: A major European city, Ingolstadt, faced challenges with passenger dissatisfaction due to inaccurate real-time bus arrival predictions. The traditional point-forecast PIS (e.g., "Bus in 7 min") often failed to account for dynamic traffic conditions and unforeseen delays, leading to frustrated commuters when the actual arrival differed from the single predicted time.
Challenge: How could the city's transport authority (VGI) leverage advanced forecasting to improve passenger trust and satisfaction by more realistically communicating arrival times, without overwhelming users with complex data?
Solution: Drawing insights from academic research, VGI implemented a pilot study comparing their existing point forecast system with a new interval-based prediction display. This new system presented arrival times as a range (e.g., "4-10 min"), reflecting the inherent uncertainty of real-world conditions. The interval widths were dynamically generated based on historical delay distributions, converging to a precise point as the bus neared the stop.
Outcome: The interval forecast system proved highly effective. When buses arrived early or late, passengers reported significantly higher satisfaction with the interval displays compared to the traditional point forecasts. While point forecasts were preferred for exact arrivals (a rare occurrence in unpredictable urban environments), the interval approach dramatically mitigated negative sentiment during delays or early arrivals. This shift demonstrated that passengers ultimately valued accuracy over misleading precision, leading to a clearer and more consistent understanding of expected arrival times.
Impact: By embracing transparent uncertainty communication, VGI can enhance passenger trust, reduce perceived waiting time anxiety, and improve the overall perceived quality of its public transport services, fostering greater ridership and loyalty.
Calculate Your Potential AI Impact
Estimate the ROI of implementing transparent AI-driven forecasting in your operations. See how improved communication and prediction accuracy can translate into tangible benefits.
Estimated Annual Savings
$0
Annual Hours Reclaimed
0
Your AI Implementation Roadmap
A structured approach to integrating advanced forecasting and communication into your public transport PIS.
Phase 1: Data Assessment & Model Adaptation
Audit existing prediction systems and historical delay data. Adapt or develop AI models (e.g., ConvLSTM) to generate interval forecasts based on real-world variability.
Phase 2: UI/UX Design & Prototyping
Design and prototype user interfaces for displaying interval forecasts. Conduct iterative user testing to optimize clarity and ensure transparent communication of uncertainty.
Phase 3: Pilot Deployment & A/B Testing
Deploy interval forecasts in a limited pilot program. Conduct A/B tests against traditional point forecasts, measuring key metrics like passenger satisfaction and perceived accuracy.
Phase 4: Full-Scale Integration & Monitoring
Roll out the enhanced PIS across all relevant channels. Continuously monitor model performance, user feedback, and make data-driven adjustments for ongoing optimization.
Ready to Elevate Your Public Transport PIS?
The future of public transport relies on transparent, accurate passenger information. Connect with our AI experts to design and implement a forecasting solution that builds trust and enhances user experience.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat