An agent with a path of her own
The proposal gives Baby Velia an identity, goals and an activity cycle distinct from Velia’s. She might take an interest in drawing, build a short story, return to an unfinished game or choose to set an attempt aside. A person could talk to her directly and revisit what she has created over time.
The character makes these changes easier to follow: a gesture can accompany a question, uncertainty or an invitation. Personality and mood belong to the interaction design; they do not establish feelings or subjective experience.
What evolving means
The proposed evolution concerns what the agent preserves and how she approaches activities. A skill, meaning a reusable procedure for a particular task, could change after several attempts. To find out whether a change helps, we need to compare it with the previous method, using new cases and cases that already worked.
- Gain experience
- Propose a variation
- Put it to the test
- Compare outcomes
- Adopt or restore
The goal is for Baby Velia to propose, test and adopt these changes autonomously, using the tools and resources in her own environment. Extending that environment requires defining new access permissions and new tests. Keeping the previous version makes it possible to recover a method that worked.
For example, Baby Velia might experiment with a different way of composing an illustrated scene. If the criterion is keeping the character recognisable, we compare the new images against that requirement. A change in taste can remain a preference: the Lab needs to distinguish a creative choice from a measured improvement. This describes a possible experiment that has yet to be implemented.
Learning through exchange
Velia could act as a mentor, offering a lesson or discussing an attempt. Baby Velia would retain her own cycle and would need to test what she receives before adopting it. This cooperation is part of the design; two agents’ results still need checking, because they can share the same mistake.
Where the project stands
A design under review, mockups and character and animation studies are available. The autonomous cycle described here is a development goal: visual studies do not yet demonstrate learning or evolution in the product. The Lab will follow the move from hypotheses to the first experiences that can be checked.