A dimensional summation account of polymorphous category learning.

Polymorphous concepts are hard to learn, and this is perhaps surprising because they, like many natural concepts, have an overall similarity structure. However, the dimensional summation hypothesis (Milton and Wills Journal of Experimental Psychology: Learning, Memory and Cognition, 30, 407-415 2004) predicts this difficulty. It also makes a number of other predictions about polymorphous concept formation, which are tested here. In Experiment 4, we confirm the theory's prediction that polymorphous concept formation should be facilitated by deterministic pretraining on the constituent features of the stimulus. This facilitation is relative to an equivalent amount of training on the polymorphous concept itself. In further experiments, we compare the predictions of the dimensional summation hypothesis with a more general strategic account (Experiment 2), a seriality of training account (Experiment 3), a stimulus decomposition account (also Experiment 3), and an error-based account (Experiment 4). The dimensional summation hypothesis provides the best account of these data. In Experiment 5, a further prediction is confirmed-the single feature pretraining effect is eliminated by a concurrent counting task. The current experiments suggest the hypothesis that natural concepts might be acquired by the deliberate serial summation of evidence. This idea has testable implications for classroom learning.

Are paid tools worth the cost? A prospective cross-over study to find the right tool for plagiarism detection.

Background: The increasing pressure to publish research has led to a rise in plagiarism incidents, creating a need for effective plagiarism detection software. The importance of this study lies in the high cost variation amongst the available options for plagiarism detection. By uncovering the advantages of these low-cost or free alternatives, researchers could access the appropriate tools for plagiarism detection. This is the first study to compare four plagiarism detection tools and assess factors impacting their effectiveness in identifying plagiarism in AI-generated articles. Methodology: A prospective cross-over study was conducted with the primary objective to compare Overall Similarity Index(OSI) of four plagiarism detection software(iThenticate, Grammarly, Small SEO Tools, and DupliChecker) on AI-generated articles. ChatGPT was used to generate 100 articles, ten from each of ten general domains affecting various aspects of life. These were run through four software, recording the OSI. Flesch Reading Ease Score(FRES), Gunning Fog Index(GFI), and Flesch-Kincaid Grade Level(FKGL) were used to assess how factors, such as article length and language complexity, impact plagiarism detection. Results: The study found significant variation in OSI(p < 0.001) among the four software, with Grammarly having the highest mean rank(3.56) and Small SEO Tools having the lowest(1.67). Pairwise analyses revealed significant differences(p < 0.001) between all pairs except for Small SEO Tools-DupliChecker. Number of words showed a significant correlation with OSI for iThenticate(p < 0.05) but not for the other three. FRES had a positive correlation, and GFI had a negative correlation with OSI by DupliChecker. FKGL negatively correlated with OSI by Small SEO Tools and DupliChecker. Conclusion: Grammarly is unexpectedly most effective in detecting plagiarism in AI-generated articles compared to the other tools. This could be due to different softwares using diverse data sources. This highlights the potential for lower-cost plagiarism detection tools to be utilized by researchers.