With support from the NSF’s “Improving Undergraduate STEM Education” program, we developed an artificial intelligence based cognitive assistant, called sInvestigator (science Investigator) that helps students develop critical thinking skills for addressing scientific problems, through an easy to employ inquiry-based approach. The instructor formulates an inquiry and the students hypothesize possible answers. Then they use sInvestigator to determine the hypothesized answer that is best supported by evidence. They form teams, each team developing an evidence-based argumentation for assessing the probability of their selected hypothesized answer. Using sInvestigator, they decompose their hypothesis into simpler and simpler hypotheses, until the top hypothesis is reduced to very simple (elementary) hypotheses, each showing clearly what evidence may favor or disfavor it. This guides the students to search for this evidence on the Internet, and assess its relevance and credibility by using fuzzy qualifiers, such as “likely” or “almost certain”. Based on these student assessments, sInvestigator determines the probabilities of the elementary hypotheses, and combines them to obtain the probability of the hypothesized answer. The students may be asked to perform deeper credibility analyses of the Internet articles found, by assessing lower-level credentials, such as author’s competence, conflicts of interest, and publication reputation. They may also be asked to justify their assessments. Once an analysis is complete, sInvestigator generates an analysis report that is finalized by the students and debated in class. sInvestigator was experimentally used in several GMU undergraduate science classes. Case studies for middle school and high school science classes have also been developed.