Mijke the Matchbot was developed to address this challenge. Mijke is an AI-powered conversational assistant that helps people find the right support in a friendly and accessible way. Instead of searching through websites or navigating complex systems, users can simply describe their situation and receive guidance towards relevant organisations and services.

Over the past two years, Sichen Guo pursued her EngD traineeship within Eindhoven Engine’s Inclusive Society focus area. In her final year she dedicating her work to optimizing Mijke the Matchbot.

Bridging the gap between people and support

“My research focused on how conversational AI could help people with limited basic skills navigate Dutch social services,” says Sichen. “Many support systems still require people to search, compare, and interpret information independently, which can be challenging when dealing with financial, health, housing, or social issues.”

In her first year Sichen explored simplifying formal letters and service information. With these insights she was able to evaluate Mijke’s effectiveness and better understand user needs. She also focused on voice-enabled communication as a way to make the system even more accessible.

Building responsible AI

As the project progressed, the focus expanded beyond building a functional chatbot. Sichen investigated how such a system could be maintained, evaluated, and responsibly deployed over time.

Working closely with libraries, social workers, and other practitioners, she helped define five criteria for high-quality referrals:

• Relevance
• Trustworthiness
• Accuracy
• Clarity
• Completeness

 

These criteria formed the basis of a human-centered evaluation framework.

One of the most important findings was what Sichen calls the “Fluency Trap.” AI-generated responses can sound convincing, friendly, and professional, even when the information is incomplete or uncertain. This highlights the importance of continuous evaluation and human oversight when using AI in public-service settings.

Looking ahead

Looking forward, Sichen is interested in exploring European AI models and specialized AI solutions designed specifically for social support contexts. Her work shows that responsible AI requires more than advanced technology. It depends on collaboration between designers, practitioners, organizations, and users. Rather than replacing human support, conversational AI can help lower barriers and make social services more accessible for everyone.

We are delighted that Sichen will continue her work on Mijke the Matchbot at Eindhoven Engine, helping to further develop and strengthen this innovative solution.

• Contact our Inclusive Society Team for more information.

An inspiring example of learning by doing and collaboration. Rijkswaterstaat provided the experimental space, Bouwbedrijf Th. van Kasteren as part of ‘Coöperatie Symbiotisch Bouwen’ made this operationally possible, whilst Eindhoven Engine is the orchestrator of the broader circular water project for the region. It is great to see the collaboration between these partners!

𝗪𝗵𝗮𝘁’𝘀 𝗻𝗲𝘅𝘁?
The next phase will start soon, building on the same principle: capturing, filtering, storing, and reusing water. This time sourced from the highway A58 for local applications such as agriculture.

A unified approach to innovation

Coordinated and managed by Kropman B.V., the consortium includes installation companies, energy consultants, platform developers, building owners and managers, technology providers, and other experts. The aim is to foster open knowledge exchange and drive innovation through strategic collaboration and widespread dissemination of results.

Buildings at the heart of the energy transition

As the energy sector shifts towards a decentralized, digital, and low-carbon future, buildings—significant consumers of electricity and thermal energy—are key to providing demand-side flexibility. This flexibility allows buildings to adjust their energy use in response to external signals like electricity prices, carbon intensity, or grid constraints, without sacrificing occupant comfort or essential functions.

Kick-off meeting 17th June, 2025 | Kropman Nijmegen

The power of model predictive control

Model Predictive Control (MPC) is a cutting-edge tool for optimizing building energy performance. Unlike traditional control systems, MPC uses predictive models and real-time data to proactively manage a wide range of systems, including HVAC (Heating, Ventilation, and Air Conditioning), lighting, heat pumps, thermal storage, electric vehicles charging and other building loads It uses forecasts of occupancy, weather conditions and other relevant parameters to determine the best control actions over a future period, continuously updating the plan based on new data.

Focus on energy flexibility

This project brings together 12 partners, including Eindhoven Engine. It aims to enhance the reliability and efficiency of local energy supply by better utilizing energy flexibility. Additionally, this initiative is particularly crucial for large, complex building systems equipped with heat pumps and thermal energy storage. These systems often experience stability and coordination issues, which can negatively impact both energy and comfort performance. Therefore, by focusing on energy flexibility, BuildInFlexergy aims to overcome these challenges and ensure a more stable and efficient energy supply for building.

Smart control systems and model predictive control

By transforming building energy management with smart control systems and model predictive control, we achieve real-time responsiveness and dynamic climate control. Moreover, this optimization uses key performance indicators. This unified approach to innovation places buildings at the heart of the energy transition, paving the way for the future of building energy flexibility.

Addressing information inequality

Millions of people in the Netherlands struggle with basic literacy, numeracy, and digital skills. Mijke shows a possibility of solving this information inequality by offering a friendly experience tailored to the real-world needs of often overlooked people via Artificial Intelligence (AI). This work stands at the crossroads of technology, social equity, and public service innovation. It shows how AI can be used responsibly in human-centered design.

Empowering communities

To achieve these goals, I plan to run iterative user tests to gather insights and improve mijke’s usability.  to help people with limited basic skills access support more easily and confidently. By doing so, I hope to decrease the workload of stakeholders, such as social workers, employees from municipalities, or public service staff. Additionally, I want to set a benchmark for evaluating AI recommendations in social public service contexts and contribute new knowledge and tools to the fields of human-computer interaction (HCI), accessibility, and digital inclusion.