Some conversations feel like open water.

Project Harbour aims to contribute to better social and professional inclusion for everyone by assessing and developing conversational skills, regardless of a person’s abilities or profile.

Spoken conversation is a fundamental part of human interaction and an essential means of building social connections. The way we speak, interact with others and use our voice can have a major impact on the quality and outcome of a conversation. Expectations vary depending on the situation, the people involved and their cultural backgrounds. Core conversational skills, however, can be developed through practice.

As part of this study, we are currently providing access to a prototype that lets you try several use cases by practising with a virtual conversation partner.

You have a short spoken conversation with a virtual conversation partner, in complete confidentiality. Before starting, you choose a scenario from a range of situations, based on the type of conversation or the skills you would like to work on.

After each conversation, the application analyses the interaction and gives you objective, detailed feedback grounded in language sciences. Based on the strengths and the areas for improvement it identifies, it offers specific, practical recommendations you can readily apply.

Artificial intelligence (AI) is used to interact with you and to bring the virtual conversation partner to life while it follows a predefined scenario. AI is also used to spot specific elements within the conversation that have been defined in advance by the application’s programme. The core of the assessment therefore does not rest on an AI freely managing and interpreting a conversation. It rests on a programme that determines which elements are analysed and how they are assessed. This approach is designed to give accurate, reproducible and explainable results.

Improving conversational skills then depends on repeated practice and on applying the recommendations the application provides. The tool is designed to support that progress over time through a personalised learning pathway, with the aim of encouraging sustained engagement, strengthening self-confidence and reducing anxiety around conversational situations.

The application needs an internet connection. It works best on a computer in a quiet place, ideally with a headset and microphone.

It can also be used on a mobile phone, although interference may occur depending on the quality of the mobile connection and the surrounding noise.

We are working on those technical improvements. For the best experience at this stage, we recommend a computer and a headset, somewhere with a good internet connection.

Every tester must be aged 16 or older. You may test Harbour for yourself, as a parent or caregiver, or as a professional supporting other people.

The age of someone you support does not determine your own eligibility. The screening questions ask separately about your age eligibility and the people you have in mind.

The application is available in English and French. You need to speak the language of the version you want to use fluently, whatever your native language or where you live.

To reflect the diversity of potential users, places are limited within each category, such as language or intended use. Screening confirms whether a place is available for you.

New study registrations are open through 30 November 2026 (Europe/Zurich), subject to available places.

Entirely. You may withdraw at any moment, without giving an explanation and without any consequences.

For this pilot test we collect your name, place of residence and email address, in order to record your agreement and to send you reminders about your progress through the test.

Separately, and as a core function of the application, we record your voice and speech during practice sessions (approximately 10 minutes per session). Those recordings are pseudonymised: they are stored under a code, not your name.

We also collect your views on the pilot test through in-app questionnaires. Those answers are pseudonymised too, and are not linked to your identity.

Your voice recordings and questionnaire answers are stored and accessed entirely separately from your personal contact details, pseudonymised, on servers in the European Union and Switzerland. Your account data — your name, place of residence, email address and the record of your consent — is processed by Google Cloud and by our email provider Resend, under data-processing agreements. We never sell your data, and we never share it with third parties beyond those contracted service providers, or where the law requires it.

Only our Chief Technology Officer (CTO) can, in exceptional circumstances — for example where the law requires it — establish a link between the two. Your personal data is kept for a maximum of five years.