Schemas are known to facilitate learning and memory consolidation in humans and rodents. However, their role in decision making is not well understood and the effects of task factors like risk, novelty, prior knowledge, arousal and attention are under-explored. We designed a new computer-based experiment where participants could firstly explore a list of different author’s paintings or quotes and then select items with the same authorship. Most participants showed improvement in performance, yet factors, including previous progress, risk and novelty conditions, pupil diameter and heart rate, prior knowledge, English proficiency and task type were linked to different aspects of their performance. Previous progress encouraged the participants switch from choosing low payoff items to high payoff items, leading to further improvement. Eye tracking analysis revealed that participants firstly screened all items to choose a schema, and then selected the matching items, shown by the difference in metrics like response times or numbers of observed items. We also designed a computational model based on error-based learning and drift diffusion processes. In our model, choice options are integrated and selected based on schema confidence (which is affected by exploration and performance feedback of that schema), payoffs for different schemas, and attentional factors. Model simulation replicated the observed patterns in choice behaviour and various cognitive measures. The estimated parameters revealed notable differences between paintings and quotes tasks. We show that schemas assist learning and decision making with the modulation of individual differences, biomarker levels and environmental factors, which interact in a complex manner.