[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-82069-en":3,"doc-seo-82069-105":29,"detail-sidebar-cat-0-en-105":90},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},82069,13056703019404,"Miles","https://ap-avatar.wpscdn.com/davatar_29158cc5080c5b710cf443261637dec0",8,"Research & Report","Experimental Evidence on the Learning Impact of Generative AI","Randomized in-person experiment testing how generative AI access changes student learning. Undergraduates study an unfamiliar technical topic and write an analytical essay either with or without off-the-shelf generative AI, then complete unaided knowledge tests and essays immediately and one week later. AI access increases immediate knowledge-test scores by 0.27 standard deviations and the gains persist. Essay quality shifts are smaller short-run, but improve in style and relevance after one week. Delayed benefits are larger for augmentation users who use AI to explain concepts, not generate text. Mechanisms include reallocating time toward reading/searching and reporting higher learning enjoyment.","arXiv :2607 .08849v1 [ econ .GN] 9 Jul 2026  \nExperimental Evidence on the Learning Impact of  \nGenerative AI ∗  \nZara Contractor Germán Reyes  \nJuly 2026  \nAbstract  \nWe study how generative AI affects student learning in a randomized experiment. In proctored, in-person sessions, undergraduates learn about an unfamiliar topic and write an analytical essay with or without access to off-the-shelf generative AI, then complete unaided assessments immediately and one week later. We measure learning with knowledge tests (factual and conceptual understanding) and openended essays (higher-order skills) . AI access raises immediate test scores by 0.27 standard deviations. These gains persist one week later. Essay quality, by contrast, changes little while students have AI access but improves in style and relevance one week later, when students write unaided. These delayed gains are larger among augmentation users—who use AI to explain concepts rather than generate text—whereas automation users’ short-run quality gains vanish once AI is removed. We find evidence for two mechanisms behind the learning gains: students shift time away from drafting text and toward reading and searching for information, and they report greater learning enjoyment.  \n∗ Contractor: Middlebury College ([zcontractor@middlebury.edu](zcontractor@middlebury.edu)), Reyes: Middlebury College and IZA ([greyes@middlebury.edu](greyes@middlebury.edu)). For helpful discussions and comments, we thank Chris Campos, Jeff Carpenter, Amy Collier, Sam Hirshman, David Huffman, Guy Ishai, Cory Koedel, Seunghoon Lee, Peter Matthews, David Munro, Caitlin Myers, Ted O’Donoghue, Evan Riehl, Andrea Robbett, Nick Swanson, and participants at the Liberal Arts Colleges Behavioral and Experimental Economics Workshop (LACBEE), the NBER Economics of Education Spring Meeting, the New England Experimental Economics Workshop, the Northeastern Economics of Education Workshop, the Royal Economic Society Conference, the Society of Labor Economists Annual Meeting, and seminars at Cornell University, Middlebury College (Carol Rifelj Faculty Lecture), Ohio University, the University of Missouri, and the University of Texas Rio Grande Valley. Financial support from Middlebury College’s Office of the Provost is gratefully acknowledged. We thank Kimberly Barros, Rebecca Deranian, Brooke Dolan, Alice Gindin, Sarah Hayward, Ember Pikramenos, Kevin Ramirez, Ephraim Shinko, and Juan Marcelo Verdugo for their assistance with experiment proctoring. The study was reviewed and approved by the Institutional Review Board of Middlebury College. Survey instruments and additional materials are available in the Supplementary Materials.  \n1 Introduction  \nGenerative artificial intelligence (AI) is now widely used in higher education. ChatGPThas over 900 million weekly users, college-age users send nearly half of its messages, and more than a quarter of those messages relate to learning (Chatterji et al. , 2025 ; OpenAI, 2025) . Despite this widespread use, evidence on whether AI helps students learn is scarce. Much of the available evidence comes from workplace settings, in which a growing literature documents large productivity gains from AI access (e.g. , Noy and Zhang, 2023 ; Brynjolfssonet al. , 2025) . However, these studies measure job performance and not human-capital accumulation. AI tools could enhance learning by acting as on-demand tutors that explain difficult concepts and provide personalized feedback (Nickow et al. , 2024), or undermine learning by offloading the cognitive effort that learning requires (Risko and Gilbert, 2016) . Whether generative AI builds or erodes human capital remains an open empirical question.  \nWe study whether off-the-shelf AI affects immediate and longer-term learning in a common class of academic tasks: time-bounded assignments in which students must learn unfamiliar material, identify relevant information, and demonstrate understanding through writing and tests. Our randomi","cbCaij1FXww30fkL","https://ap.wps.com/l/cbCaij1FXww30fkL","pdf",10973079,1,93,"English","en",105,"# Abstract\n# Introduction","[{\"question\":\"How does the study test the learning impact of generative AI?\",\"answer\":\"The study runs a randomized experiment with proctored, in-person sessions. Students learn an unfamiliar topic, write an analytical essay with or without access to off-the-shelf generative AI, and then complete unaided assessments immediately and one week later.\"},{\"question\":\"What learning outcomes improve when students can use generative AI?\",\"answer\":\"AI access raises immediate knowledge-test scores by 0.27 standard deviations, and these gains persist one week later. Higher-order outcomes measured through open-ended essays show different timing and patterns of change.\"},{\"question\":\"Do all AI users benefit the same way?\",\"answer\":\"No. Delayed gains are larger for augmentation users who use AI to explain concepts rather than generate text, while automation users’ short-run quality gains disappear once AI access is removed.\"}]",1784178022,234,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":85,"head_meta":87,"extra_data":89,"updated_unix":27},"experimental-evidence-on-the-learning-impact-of-generative-ai","",{"@graph":35,"@context":84},[36,53,67],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/experimental-evidence-on-the-learning-impact-of-generative-ai/82069/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":61,"encodingFormat":60,"isAccessibleForFree":62,"interactionStatistic":63},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-16",true,{"@type":64,"interactionType":65,"userInteractionCount":4},"InteractionCounter",{"@type":66},"ViewAction",{"@type":68,"mainEntity":69},"FAQPage",[70,76,80],{"name":71,"@type":72,"acceptedAnswer":73},"How does the study test the learning impact of generative AI?","Question",{"text":74,"@type":75},"The study runs a randomized experiment with proctored, in-person sessions. Students learn an unfamiliar topic, write an analytical essay with or without access to off-the-shelf generative AI, and then complete unaided assessments immediately and one week later.","Answer",{"name":77,"@type":72,"acceptedAnswer":78},"What learning outcomes improve when students can use generative AI?",{"text":79,"@type":75},"AI access raises immediate knowledge-test scores by 0.27 standard deviations, and these gains persist one week later. Higher-order outcomes measured through open-ended essays show different timing and patterns of change.",{"name":81,"@type":72,"acceptedAnswer":82},"Do all AI users benefit the same way?",{"text":83,"@type":75},"No. Delayed gains are larger for augmentation users who use AI to explain concepts rather than generate text, while automation users’ short-run quality gains disappear once AI access is removed.","https://schema.org",{"og:url":51,"og:type":86,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":88,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":91},[92,96,100,104,109,114,119,122,127,130,134],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":93,"show_sort_weight":94,"slug":95},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":97,"show_sort_weight":98,"slug":99},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":101,"show_sort_weight":102,"slug":103},"Exam",70,"exam",{"id":105,"doc_module":4,"doc_module_name":45,"category_name":106,"show_sort_weight":107,"slug":108},5,"Comic",60,"comic",{"id":110,"doc_module":4,"doc_module_name":45,"category_name":111,"show_sort_weight":112,"slug":113},6,"Technology",50,"technology",{"id":115,"doc_module":4,"doc_module_name":45,"category_name":116,"show_sort_weight":117,"slug":118},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":45,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":45,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":45,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":45,"category_name":136,"show_sort_weight":105,"slug":137},19,"General","general"]