Not Whether but What: A Systematic Review of the Locus of Offloading in Generative AI and Higher-Order Thinking
INTERNATIONAL JOURNAL OF HUMAN-COMPUTER INTERACTION, cilt.42, sa.22, ss.1-20, 2026 (SCI-Expanded, SSCI, Scopus)
- Yayın Türü: Makale / Derleme
- Cilt numarası: 42 Sayı: 22
- Basım Tarihi: 2026
- Doi Numarası: 10.1080/10447318.2026.2763152
- Dergi Adı: INTERNATIONAL JOURNAL OF HUMAN-COMPUTER INTERACTION
- Derginin Tarandığı İndeksler: Applied Science & Technology Source, Academic Search Ultimate (EBSCO), Business Source Ultimate (EBSCO), Engineering Source (EBSCO), Scopus, Psychology & Behavioral Sciences Collection (EBSCO), Technology Collection (ProQuest), Aerospace Database, Science Citation Index Expanded (SCI-EXPANDED), Social Sciences Citation Index (SSCI), ABI/INFORM, Compendex, INSPEC, Psycinfo
- Sayfa Sayıları: ss.1-20
- Bursa Uludağ Üniversitesi Adresli: Evet
Özet
Generative artificial intelligence (GenAI) reduces the effort of learning tasks, but its effect on students’ higher-order thinking may depend on what is offloaded. This pre-registered systematic review synthesized 26 studies (2024–2026), identified from 4,729 records, that measured cognitive load and higher-order thinking in the same higher education sample. Across 21 experimental studies, GenAI conditions most often showed lower load (12 studies) and better higher-order outcomes (14 studies). Lower load coincided with poorer thinking in three studies, none of high quality, and in all three the tool was positioned to perform the target thinking. Most outcomes were self-reports or products created using the available tool, and the only study assessing thinking after the tool was withdrawn found a decline. We propose, as a hypothesis, that offloading supports learning when the tool takes over preparatory work but may undermine it when the tool takes over the thinking the task should exercise.