In anticipation of conceptualising a co-creation lab, a demonstrator (technical software application) was built under the guidance of DFKI as technical partner and in cooperation with social sciences in order to better understand procedural needs. It turned out to be a highly participative and complex task to identify requirements for building better technology by societal interaction in a future co-creation lab. A social assessment system was designed and developed on basis of a neural network, which performs assessment practices (based on anonymised census data as training data) that will potentially be used in future social welfare systems. The interactive design process was then reviewed upon and enriched by sociological knowledge and concepts in order to inform the concept of a co-creation lab during the ongoing project.

Designing an AI-based social assessment system

From a technical point of view, an assessment process has three main steps:

  1. Obtain process-relevant data
  2. Profile data
  3. Assess / decide based on calculated profiles

If the profiling is done with machine learning, such a system will probably have the following components:

  • Sample data for learning/training on profiles
  • Live data for assessment
  • Profiler module profiling live data (e.g. calculating probability predictions for each profile)
  • Rule-based decision system for final decision making regarding profiles

The process needs to include checks that continuously revise the built-in rules and the decisions of the system to ensure its compliance with ethical norms and laws. To specify these checks will be task of the next steps in the AI FORA project.