I see the importance of "learning purposes" as a guide to how the assessment steps can be planned out. If a learner can identify a learning purpose(s) at each study/lesson, perhaps, the system might be able to find a closer match as well. I think this could help with your assessment targets.
(Alignment of : Learning outcomes/purposes - Learning Activities - Assessments)
In the virtual world, how can one be certain with the accuracy/reliability of the stories? In a supervised environment, this can be verified. If the learning in your innovation does not care about the accuracy of the contents, only use the contents as part of the learning process, a sentence to stress this may clear such concern.
As for lightweight experiment, a 'physical' pilot test can be beneficial to test out the idea. This can be done in a classroom in conjunction with a Creative Writing lesson where students mimic your proposed steps, - write their stories - compile their profiles - make story boards - initiate 'interest/experience/learning purpose market' where story provide and seekers meet, match and share content - follow the rest of the steps as in your proposal emphasizing on how to install "effective and on-target" learning assessment processes. I think this physical platform will give valuable input into your design especially the matching step. The algorithm matching will match bits and bytes of learners' inputs but, in the real world, our minds have more nuances than bits and bytes.
Hope this helps and relevant to your project. Excited to see this project come to live!