Machine Intelligence Quotient

= Machine Intelligence Quotient =

Intelligent agent-based systems (IABSs) are a common type of intelligent systems. An IABS could be an intelligent agent (IA) which solves problems individually, or an intelligent cooperative multiagent system (ICMS) where the agents cooperate in problem-solving. The health sciences employ intelligent systems for a wide range of tasks, including to solve complex medical diagnosis problems. defines the novel idea of the metaverse in intelligent healthcare. In the context of intelligent systems, measuring problem-solving machine intelligence becomes of utmost importance. There are very few studies that present results related to the subject of measuring machine intelligence. Machine intelligence metrics presented in scientific literature rely on different philosophies, which hinders their effective comparison. There is no standardization on what machine intelligence is and what should be measured to quantify it.

Universal black-box-based machine intelligence metrics
To ensure the universality of machine intelligence metrics, a feasible general approach consists of the black-box-based intelligence modeling used to measure the central intelligence tendency of intelligent agent-based systems. Such metrics should treat aspects like the variability in intelligence, and anomalies (outliers) in intelligence measurements. Statistically grounded universal methods with these properties presented for measuring the machine intelligence quotient of intelligent agent-based systems (IABSs) consist of MetrIntPair, MetrIntPairII , MeasApplInt and MetrIntSimil. presents a guide for choosing the most appropriate black-box-based intelligence metric for measuring the intelligence of a set of studied IABSs, classification of IABSs in intelligence classes, and detection of the IABSs with statistically low and high outlier intelligence. The general idea of these intelligence measuring methods consists in the fact that each of them can be applied to a set of studied intelligent systems, whose problem-solving intelligence is intended to be measured. Initially, an intelligence measurement of the studied systems is performed on a set of difficult testing problems. After this, as a second step, the obtained intelligence measurement data is analyzed based on advanced statistical methods. As a result, the Machine Intelligence Quotient (MIQ) that represents the so-called machine intelligence tendency is calculated using advanced statistical methods. The calculated MIQs are comparable.

Pairwise measuring intelligence of agent-based systems
MetrintPair is an intelligence metric that can be applied to two studied IABSs. IABSs with statistically the same intelligence are classified in the same class. The specificity of the measurements consists in the fact that they are made pairwise (a difficult problem is chosen and the problem-solving intelligence of both systems for the respective problem is measured). A limitation of the metric consists in the fact that it can be applied to only two studied IABSs. In the case of application of more than two IABSs, a so-called family-wise error rate (FEWER) statistical error appears. To eliminate this disadvantage, MetrIntPairII intelligence metrics, was developed, having the same properties as MetrintPair; it can be applied to any number of IABSs.

Non-pairwise measuring intelligence of agent-based systems
MeasApplInt is a similar intelligence metric to MetrintPair, being applicable to two studied IABSs. The applicative difference consists in the fact that the experimental problem-solving evaluation does not require pairwise problem-solving intelligence measurements. MetrIntSimil extends MeasApplInt, being applicable to intelligence measuring of any number of intelligent systems. The systems with the same intelligence are classified in the same class. The difference between MetrIntSimil versus MetrIntPairII consists of the specific statistical data analysis since the experimental problem-solving intelligence in the case of MetrIntSimil is not made pairwise.