Although the number and diversity of intelligent agent-based systems (IABSs), also known as agentic systems, evolve very rapidly, the main difficulty of measuring the intelligence of intelligent systems, also referred to as machine intelligence, remains. Previously, we have developed some novel intelligence metrics for measuring machine intelligence, namely MetrIntPairII, ExtrIntDetect, and MetrIntSimil. In this paper, the focus is on the MetrIntSimil metric. MetrIntSimil can measure the intelligence of an arbitrary set of IABSs, calculating their socalled intelligence quotients that are comparable, the IABSs with the same intelligence are classified in the same classes. MetrIntSimil is based on complex data science and statistical analysis. A difficult-to-decide use case is presented in that it studies and compares the intelligence of a set of intelligent systems using MetrIntSimil. Difficulties in the analysis consist of aspects like different well-known normality tests giving contradictory results, decision on outlier removal, decision between parametric and non-parametric statistical tests, and others. This is also illustrative to the artificial intelligence research, when the experimental results must be analysed using advanced data science and statisticalmethods.
Laszlo Barna Iantovics (Wed,) studied this question.