We've named her Assunta. A name that is a wish for assumption...

She is an Artificial assistant that allows users to engage in dialogue when seeking information about employment services. Through questions and answers, this tool enables exploration of candidates' experiences, training, expectations, and goals.

We are training Assunta on professional repertoires to define the target competencies for certification. But before that, Assunta gathers information to understand the elements necessary to assist users. It's a sort of pre-screening to frame the candidate's story and provide orientative advice. Following that, there are services that involve direct interaction with the candidate. Human, personal, and professional operators from our team are always present and involved in the research and experimentation of the Talenti Latenti project. During this phase, all the information and materials collected within the scope of the project research are provided to the artificial intelligence, which processes and compares them with other information in its possession to construct a domain of specific skills, both in terms of regulations, procedures, and services and in the field of neurodivergences.

The need for this arose during the analysis and development of the platform. The sheer volume of information related to regional, sectoral, national, and international repertoires, often with incorrect codes and references, forced the technical team to think of a solution that could distill useful summaries for service setup on one hand and engage with users on the other. This was no easy task with traditional database query tools. When references are not precise and accurate, the results are unusable.

The fundamental problem in the work initiated with the experimentation was the considerable volume of often contradictory and incomplete Repertori information. Nevertheless, there are logics connecting these pieces of information. Given the volumes of descriptive data and codes, this work can be effectively performed by artificial intelligence. Furthermore, if we consider a service that is transnational, the issue becomes even more complex.

We have chosen the most powerful engines available on the market to ensure the most effective and fast response to the inputs we needed to process. Our technological partner, Wiki Holding, has provided the software, and the CEQF Project Team is carrying out specific training activities using research materials and others supplied by partners.

The results are very promising. But what should users do?
Engage in dialogue with simple questions and answers. The simplest ones.

The interface consists of a simple chat that allows the collection and cataloging of information. Questions and answers highlight elements useful for the orientation counselor, certification manager, expert, and user in setting up an evaluation, training, and employment placement program that takes into account both formal and substantive aspects of the context.

In this initial phase, we are feeding the system with information mainly concerning Italy, and then we will incorporate what we've gathered in Spain and France.

Artificial intelligences use Learning Language Models (LLMs) to process information and respond to the inputs they collect, for example, during training and user interactions. As a result, there isn't a "defined result" but rather a process that generates connections, much like in a normal conversation. The more information there is, the more connections grow, and the responses become more elaborate and personalized.

It will be very interesting to see how the interaction with AI is evaluated by a neurodivergent person. Assunta's attitude will also be important, and it depends on the training we are providing in this regard.

Stay tuned for updates.

TalentiLatenti Project