01
Model user profiles
Describe target users as first-class models: language, accessibility needs, mobility constraints, competences, and preferences.
Features · Personalization
Personalized agents start with explicit user models.
BESSER turns personalization into a visible low-code pipeline. You model who your users are, configure how the agent should adapt, validate the result, and generate profile-aware conversational behavior without writing the personalization logic by hand.
02
Configure adaptations
Define which parts of the agent should adapt, from presentation and modality to content and behavior.
03
Review the variant
Inspect the personalized result, refine generated text, and validate the adapted model before shipping it.
04
Generate and run
Produce personalized agent behavior without hand-written personalization logic in the final application.
User modeling
Start by modeling the user, not by hard-coding assumptions.
The pipeline begins with an explicit user profile model. Instead of leaving accessibility constraints, language preferences, or competences in informal notes, BESSER represents them as structured modeling elements that can later drive automatic adaptations.
- Graphical user modeling in the editor instead of buried requirements text
- Profiles can capture disabilities, cultural aspects, competences, language, and preferences
- One profile becomes reusable input for both interface and agent personalization
Agent customization
Then configure how the agent should adapt to that profile.
BESSER exposes personalization decisions in a dedicated configuration view. Designers can adapt language, style, readability, modality, and profile-aware content so the generated conversational agent reflects the intended experience before any code is produced.
- Adapt presentation, modality, content, and behavior from a single customization view
- Apply design-time personalization first to improve runtime performance and inspect results early
- Keep the human in the loop with manual validation and final refinements before generation
Personalized result
The outcome is not a cosmetic tweak. It changes what the agent says.
Personalization combines user profiling with system adaptation. In the gym-assistant example from the paper, the default answer suggests generic exercises, while the personalized variant adapts the content to the user's mobility needs and presents a safer, more appropriate response.
- The same base agent can speak differently depending on the modeled user profile
- Recommendations can be made accessible and context-aware, not just stylistically rewritten
- Personalization stays explicit, low-code, and reproducible across profiles
Why this matters
A low-code pipeline for personalized conversational agents.
The key idea is to separate user profiles, base agents, and agent adaptations so each artifact stays reusable. BESSER then applies those models across design-time and runtime personalization instead of relying on one-off prompt engineering or ad-hoc application code.
Reusable user profiles
A user model is a durable artifact that can be reused across interfaces, agents, and future adaptations.
Visible adaptation logic
Content, style, readability, and modality choices stay inspectable in the model instead of disappearing into implementation details.
Generated personalized behavior
The final agent can adapt to specific profiles while remaining part of the same model-driven BESSER workflow.
Walkthrough
See the full personalization flow.
This walkthrough shows the complete low-code pipeline: modeling the user, configuring the personalization, and observing the adapted agent behavior in action.
BESSER personalization demo