AI\u8f85\u52a9 vs \u4eba\u7c7b\u72ec\u7acb\u89e3\u51b3\u95ee\u9898\u6548\u7387\u5bf9\u6bd4\u2014\u2014\u6587\u732e\u7efc\u8ff0<\/title><\/p>\n<style>\n * { box-sizing: border-box; margin: 0; padding: 0; }\n body { font-family: 'Times New Roman', Georgia, serif; line-height: 1.8; color: #222; background: #fafafa; }\n .container { max-width: 900px; margin: 0 auto; padding: 40px 30px; background: #fff; }\n h1 { font-size: 1.8em; text-align: center; margin-bottom: 8px; color: #1a1a1a; }\n h2 { font-size: 1.3em; margin: 40px 0 16px; border-bottom: 2px solid #333; padding-bottom: 6px; color: #1a1a1a; }\n h3 { font-size: 1.1em; margin: 24px 0 10px; color: #333; }\n p { margin: 12px 0; text-align: justify; }\n ul, ol { margin: 10px 0 10px 30px; }\n li { margin: 6px 0; }\n .highlight { background: #fff3cd; border-left: 4px solid #ffc107; padding: 12px 16px; margin: 16px 0; }\n .box { background: #f0f4ff; border-left: 4px solid #4a90d9; padding: 12px 16px; margin: 16px 0; }\n .meta { text-align: center; color: #666; font-size: 0.9em; margin-bottom: 32px; }\n .ref { padding-left: 30px; text-indent: -30px; margin: 8px 0; font-size: 0.95em; }\n .ref-section { margin-top: 40px; border-top: 1px solid #ccc; padding-top: 20px; }\n table { width: 100%; border-collapse: collapse; margin: 16px 0; font-size: 0.9em; }\n th, td { border: 1px solid #ddd; padding: 10px 12px; text-align: left; }\n th { background: #4a90d9; color: #fff; font-weight: bold; }\n tr:nth-child(even) { background: #f8f8f8; }\n .tag { display: inline-block; padding: 2px 8px; border-radius: 3px; font-size: 0.8em; font-weight: bold; margin-right: 4px; }\n .tag-pos { background: #d4edda; color: #155724; }\n .tag-neg { background: #f8d7da; color: #721c24; }\n .tag-mix { background: #fff3cd; color: #856404; }\n @media (max-width: 600px) {\n .container { padding: 20px 15px; }\n h1 { font-size: 1.4em; }\n }\n<\/style>\n<p><\/head><br \/>\n<body><\/p>\n<div class=\"container\">\n<h1>AI\u8f85\u52a9\u4e0e\u4eba\u7c7b\u72ec\u7acb\u89e3\u51b3\u95ee\u9898\u6548\u7387\u5bf9\u6bd4<\/h1>\n<h1 style=\"font-size:1.2em;font-weight:normal;color:#555;margin-bottom:20px;\">\u2014\u2014\u805a\u7126\u590d\u6742\/\u5f00\u521b\u6027\u95ee\u9898\u7684\u6587\u732e\u7efc\u8ff0<\/h1>\n<div class=\"highlight\">\n<strong>\u6838\u5fc3\u95ee\u9898\uff1a<\/strong>\u5bf9\u4e8e\u56f0\u96be\u7684\u5f00\u521b\u6027\u95ee\u9898\uff08\u5982\u5b9a\u5236\u4e0b\u8f7d\u534f\u8bae\u3001\u5e26\u5bbd\u901f\u7387\u9650\u5236\u7b49\uff09\uff0c\u4eba\u7c7b\u5229\u7528AI\u8f85\u52a9\u4e0e\u901a\u8fc7\u72ec\u7acb\u6587\u732e\u7814\u7a76\u4e24\u79cd\u65b9\u5f0f\uff0c\u89e3\u51b3\u95ee\u9898\u7684\u6548\u7387\u662f\u5426\u5b58\u5728\u663e\u8457\u5dee\u5f02\uff1f\n<\/div>\n<p><\/p>\n<h2>1. \u5143\u5206\u6790\uff1a\u4eba+AI\u4f55\u65f6\u6709\u7528\uff1f<\/h2>\n<p>\u8fc4\u4eca\u4e3a\u6b62\u6700\u5168\u9762\u7684\u4eba\u673a\u534f\u4f5c\u5143\u5206\u6790\u7531MIT Sloan\u56e2\u961f\u5b8c\u6210\uff0c\u8986\u76d6106\u9879\u5b9e\u9a8c\uff08370\u4e2a\u6548\u5e94\u91cf\uff09\uff0c\u53d1\u8868\u4e8e <em>Nature Human Behaviour<\/em>\uff08Vaccaro et al., 2024\uff09\u3002<\/p>\n<ul>\n<li><strong>\u603b\u4f53\u7ed3\u8bba\uff1a<\/strong>\u4eba+AI\u7ec4\u5408<em>\u5e73\u5747<\/em>\u8868\u73b0\u4f18\u4e8e\u7eaf\u4eba\u5de5\uff0c\u4f46\u4e0d\u5982\u7eafAI\u7cfb\u7edf\u7684\u6700\u4f73\u8868\u73b0\u3002\u672a\u53d1\u73b0\"\u4eba\u673a\u534f\u540c\u6548\u5e94\"\uff08\u5373\u7ec4\u5408\u8868\u73b0\u8d85\u8d8a\u5404\u81ea\u6700\u4f18\u503c\uff09\u3002<\/li>\n<li><strong>\u51b3\u7b56\u7c7b\u4efb\u52a1\uff1a<\/strong>\uff08\u6df1\u5ea6\u4f2a\u9020\u68c0\u6d4b\u3001\u533b\u7597\u8bca\u65ad\u3001\u9700\u6c42\u9884\u6d4b\u7b49\uff09\u4eba+AI\u7ec4\u5408\u5e38<em>\u4e0d\u5982<\/em>\u7eafAI\u3002<\/li>\n<li><strong>\u521b\u9020\u6027\u4efb\u52a1\uff1a<\/strong>\uff08\u6587\u672c\u603b\u7ed3\u3001\u95ee\u7b54\u804a\u5929\u3001\u56fe\u50cf\u751f\u6210\u3001\u65b0\u5185\u5bb9\u521b\u4f5c\u7b49\uff09\u4eba+AI\u7ec4\u5408\u5e38<em>\u8d85\u8d8a<\/em>\u5404\u81ea\u6700\u4f73\u6c34\u5e73\u3002<\/li>\n<li><strong>\u6838\u5fc3\u6d1e\u5bdf\uff1a<\/strong>\"\u6709\u6548\u6027\u4e0d\u5728\u4e8e\u4efb\u4f55\u4e00\u65b9\u7684\u57fa\u7ebf\u8868\u73b0\uff0c\u800c\u5728\u4e8e\u4e8c\u8005\u5982\u4f55\u534f\u4f5c\u548c\u4e92\u8865\u3002\"<\/li>\n<\/ul>\n<p><\/p>\n<h2>2. AI\u8f85\u52a9\u635f\u5bb3\u72ec\u7acb\u89e3\u9898\u80fd\u529b\u2014\u2014\u56e0\u679c\u8bc1\u636e<\/h2>\n<h3>2.1 \u4ec510\u5206\u949fAI\u8f85\u52a9\u5373\u9020\u6210\u663e\u8457\u635f\u4f24<\/h3>\n<p>Liu et al. (2025) \u901a\u8fc73\u9879RCT\uff08N=1,222\uff09\uff0c\u63d0\u4f9b\u4e86\u4e00\u5957\u5b8c\u6574\u7684\u56e0\u679c\u8bc1\u636e\u94fe\uff1a<\/p>\n<ul>\n<li><strong>\u5b9e\u9a8c1\uff08N=354\uff09\uff1a<\/strong>\u6570\u5b66\u89e3\u9898\u4efb\u52a1\u3002AI\u8f85\u52a9\u7ec4\u5728\u64a4\u6389AI\u540e\uff0c\u72ec\u7acb\u89e3\u9898\u6b63\u786e\u7387\u4ece73%\u964d\u81f357%\uff08Cohen's <em>d<\/em> = \u22120.42\uff09\uff0c\u653e\u5f03\u7387\u4ece11%\u5347\u81f320%\u3002<\/li>\n<li><strong>\u5b9e\u9a8c2\uff08N=667\uff09\uff1a<\/strong>\u590d\u73b0\u5e76\u6392\u9664\u6df7\u6dc6\u53d8\u91cf\uff08\u589e\u52a0\u524d\u6d4b\u3001\u7edf\u4e00\u754c\u9762\uff09\uff0c\u6548\u5e94\u590d\u73b0\uff08\u6b63\u786e\u738771% vs 77%\uff09\u3002<\/li>\n<li><strong>\u5b9e\u9a8c3\uff08N=201\uff09\uff1a<\/strong>\u9605\u8bfb\u7406\u89e3\u4efb\u52a1\uff08SAT\u98ce\u683c\uff09\u3002\u540c\u7b49\u6548\u5e94\u590d\u73b0\uff08\u6b63\u786e\u738776% vs 89%\uff0c<em>d<\/em> = \u22120.42\uff09\uff0c\u8bc1\u660e\u6548\u5e94\u5177\u6709\u8de8\u9886\u57df\u6cdb\u5316\u6027\u3002<\/li>\n<li><strong>\u5173\u952e\u53d1\u73b0\uff1a<\/strong>61%\u7684\u7528\u6237\u76f4\u63a5\u5411AI\u7d22\u8981\u7b54\u6848\uff0c\u8fd9\u90e8\u5206\u4eba\u53d7\u635f\u6700\u4e25\u91cd\uff1b<em>\u7528AI\u8981\u63d0\u793a\/\u6f84\u6e05\u7684\u7528\u6237\u4e0e\u5bf9\u7167\u7ec4\u65e0\u663e\u8457\u5dee\u5f02<\/em>\u3002<\/li>\n<\/ul>\n<h3>2.2 AI\u52a0\u901f\u6280\u80fd\u9000\u5316<\/h3>\n<p>Hohenstein et al. (2024) \u4ece\u7406\u8bba\u89d2\u5ea6\u5206\u6790\uff1aAI\u52a9\u624b\u53ef\u80fd\u52a0\u901f\u4e13\u5bb6\u7684\u6280\u80fd\u9000\u5316\uff0c\u963b\u788d\u65b0\u624b\u7684\u6280\u80fd\u4e60\u5f97\uff0c\u4e14\u7528\u6237\u5f80\u5f80\u610f\u8bc6\u4e0d\u5230\u8fd9\u4e9b\u8d1f\u9762\u5f71\u54cd\u3002<\/p>\n<h3>2.3 \u6280\u80fd\u5f62\u6210\u7684\u5b9e\u9a8c\u8bc1\u636e<\/h3>\n<p>Shen & Tamkin (2026) \u752852\u540d\u4e13\u4e1a\u5f00\u53d1\u8005\u5b66\u4e60\u65b0\u5f02\u6b65\u7f16\u7a0b\u5e93\uff08Python Trio\uff09\u7684RCT\u8bc1\u660e\uff1aAI\u7ec4\u7684\u6982\u5ff5\u7406\u89e3\u3001\u4ee3\u7801\u9605\u8bfb\u548c\u8c03\u8bd5\u80fd\u529b\u5747\u663e\u8457\u53d7\u635f\uff0c\u4e14<em>\u5e73\u5747\u6548\u7387\u5e76\u65e0\u663e\u8457\u63d0\u5347<\/em>\u3002\u5b8c\u5168\u59d4\u6258AI\u7f16\u7801\u7684\u7528\u6237\u6709\u8f7b\u5fae\u751f\u4ea7\u529b\u6539\u5584\uff0c\u4f46\u4ee5\u5b8c\u5168\u672a\u5b66\u4f1a\u8be5\u5e93\u4e3a\u4ee3\u4ef7\u3002\u7814\u7a76\u8005\u8bc6\u522b\u51fa<strong>6\u79cdAI\u4ea4\u4e92\u6a21\u5f0f\uff0c\u5176\u4e2d3\u79cd\u4fdd\u6301\u8ba4\u77e5\u53c2\u4e0e\u5ea6\u7684\u6a21\u5f0f\u4fdd\u7559\u4e86\u5b66\u4e60\u6548\u679c<\/strong>\u3002<\/p>\n<h3>2.4 AI\u8f85\u52a9\u5bf9\u6279\u5224\u6027\u601d\u7ef4\u7684\u5f71\u54cd<\/h3>\n<p>Lee et al. (2025) \u5bf9\u77e5\u8bc6\u578b\u5de5\u4f5c\u8005\u7684\u8c03\u67e5\u53d1\u73b0\uff1aGenAI\u5e2e\u52a9\u77e5\u8bc6\u578b\u5de5\u4f5c\u8005\u642d\u5efa\u590d\u6742\u4efb\u52a1\u6846\u67b6\u5e76\u81ea\u52a8\u5316\u5de5\u4ef6\u521b\u5efa\uff0c\u4f46\u4e5f\u5bfc\u81f4\u81ea\u6211\u62a5\u544a\u7684<em>\u8ba4\u77e5\u52aa\u529b\u51cf\u5c11<\/em>\u548c<em>\u4fe1\u5fc3\u6548\u5e94\u53d8\u5316<\/em>\u3002<\/p>\n<p><\/p>\n<h2>3. \u5927\u89c4\u6a21\u73b0\u573a\u5b9e\u9a8c<\/h2>\n<h3>3.1 \"\u952f\u9f7f\u72b6\u6280\u672f\u524d\u6cbf\"\u2014\u2014BCG\/HBS\u7814\u7a76<\/h3>\n<p>Dell'Acqua et al. (2025) \u5728Boston Consulting Group\u5bf9758\u540d\u987e\u95ee\u8fdb\u884cRCT\uff0c\u53d1\u8868\u4e8e <em>Organization Science<\/em>\uff1a<\/p>\n<table>\n<tr>\n<th>\u4efb\u52a1\u7c7b\u578b<\/th>\n<th>AI\u5bf9\u6027\u80fd\u7684\u5f71\u54cd<\/th>\n<th>\u6548\u679c\u91cf<\/th>\n<\/tr>\n<tr>\n<td>\u524d\u6cbf\u5185\u4efb\u52a1\uff08\u521b\u610f\u3001\u5206\u6790\u3001\u5199\u4f5c\u3001\u8bf4\u670d\uff09<\/td>\n<td>\u663e\u8457\u63d0\u5347<span class=\"tag tag-pos\">\u2191<\/span><\/td>\n<td>\u901f\u5ea6+25%\uff0c\u8d28\u91cf+40%\uff0c\u5b8c\u6210\u7387+12%<\/td>\n<\/tr>\n<tr>\n<td>\u524d\u6cbf\u5916\u4efb\u52a1\uff08\u590d\u6742\u6218\u7565\u51b3\u7b56\uff09<\/td>\n<td>\u663e\u8457\u964d\u4f4e<span class=\"tag tag-neg\">\u2193<\/span><\/td>\n<td>\u6b63\u786e\u65b9\u6848\u4ea7\u51fa\u7387<strong>\u964d\u4f4e19%<\/strong><\/td>\n<\/tr>\n<\/table>\n<p>\u7814\u7a76\u8bc6\u522b\u51fa\u4e24\u79cd\u6210\u529f\u4f7f\u7528\u6a21\u5f0f\uff1a<strong>\"\u534a\u4eba\u9a6c\"\uff08Centaurs\uff09<\/strong>\u2014\u2014\u4eba\u673a\u5206\u5de5\u534f\u4f5c\uff1b<strong>\"\u8d5b\u535a\u683c\"\uff08Cyborgs\uff09<\/strong>\u2014\u2014\u6df1\u5ea6\u878d\u5408\u5de5\u4f5c\u6d41\u3002<\/p>\n<h3>3.2 \u8f6f\u4ef6\u5de5\u7a0b\u5927\u89c4\u6a21RCT<\/h3>\n<p>Cui et al. (2025) \u5728Microsoft\u3001Accenture\u3001Fortune 100\u4e09\u5bb6\u4f01\u4e1a\u5bf94,867\u540d\u5f00\u53d1\u8005\u7684RCT\uff08\u53d1\u8868\u4e8e <em>Management Science<\/em>\uff09\uff1aCopilot\u7ec4\u4efb\u52a1\u5b8c\u6210\u91cf<strong>+26%<\/strong>\uff0c\u4f46\u63d0\u5347<em>\u96c6\u4e2d\u5728\u5e38\u89c4\u7f16\u7801\u4efb\u52a1<\/em>\uff0c\u5bf9\u8c03\u8bd5\u3001\u67b6\u6784\u51b3\u7b56\u7b49\u590d\u6742\u4efb\u52a1\u65e0\u663e\u8457\u5e2e\u52a9\u3002<\/p>\n<h3>3.3 \u547c\u53eb\u4e2d\u5fc3\u73b0\u573a\u5b9e\u9a8c<\/h3>\n<p>Brynjolfsson et al. (2025) \u5728 <em>Quarterly Journal of Economics<\/em> \u53d1\u8868\uff1aAI\u8f85\u52a9\u4f7f\u5ba2\u670d\u4eba\u5458\u6bcf\u5c0f\u65f6\u89e3\u51b3\u95ee\u9898\u6570\u91cf<strong>+15%<\/strong>\uff0c\u6548\u679c\u5bf9\u65b0\u5458\u5de5\u66f4\u663e\u8457\uff08+34%\uff09\uff0c\u5bf9\u9ad8\u6280\u80fd\u5458\u5de5\u5f71\u54cd\u6709\u9650\u3002<\/p>\n<h3>3.4 \u4e13\u4e1a\u5199\u4f5c\u4efb\u52a1<\/h3>\n<p>Noy & Zhang (2023) \u5728 <em>Science<\/em> \u53d1\u8868\uff1a\u4e2d\u7b49\u6c34\u5e73\u4e13\u4e1a\u5199\u4f5c\u4efb\u52a1\u4e2d\uff0cChatGPT\u663e\u8457\u63d0\u5347\u4e86\u751f\u4ea7\u529b\uff08\u4efb\u52a1\u5b8c\u6210\u65f6\u95f4\u51cf\u5c11\uff09\u548c\u8d28\u91cf\u3002<\/p>\n<p><\/p>\n<h2>4. AI\u7f16\u7a0b\u52a9\u624b\u4e13\u9879\u7814\u7a76<\/h2>\n<p>GitHub Copilot\u76f8\u5173\u7814\u7a76\u6784\u6210\u4e86\u5355\u4e00\u5de5\u5177\u7684\u6700\u5927\u8bc1\u636e\u96c6\uff1a<\/p>\n<table>\n<tr>\n<th>\u7814\u7a76<\/th>\n<th>\u8bbe\u8ba1<\/th>\n<th>\u5173\u952e\u53d1\u73b0<\/th>\n<\/tr>\n<tr>\n<td>Peng et al. (2023)<\/td>\n<td>\u5bf9\u7167\u5b9e\u9a8c\uff08HTTP\u670d\u52a1\u5668\u4efb\u52a1\uff09<\/td>\n<td>Copilot\u7ec4\u4efb\u52a1\u5b8c\u6210\u65f6\u95f4<strong>\u51cf\u5c1155.8%<\/strong><\/td>\n<\/tr>\n<tr>\n<td>Cui et al. (2025)<\/td>\n<td>3\u9879RCT\u51714,867\u4eba<\/td>\n<td>\u751f\u4ea7\u529b+26%\uff1b\u6548\u679c\u96c6\u4e2d\u4e8e\u5e38\u89c4\u4efb\u52a1<\/td>\n<\/tr>\n<tr>\n<td>Developer Productivity With\/Without Copilot (2025)<\/td>\n<td>\u7eb5\u5411\u6df7\u5408\u65b9\u6cd5\u7814\u7a76<\/td>\n<td>\u8c03\u78142,631\u540d\u5f00\u53d1\u8005\uff0c\u611f\u77e5\u751f\u4ea7\u529b\u4e0e\u5b9e\u9645\u4f7f\u7528\u6b63\u76f8\u5173<\/td>\n<\/tr>\n<tr>\n<td>GitHub Copilot: Asset or Liability? (2023)<\/td>\n<td>\u5bf9\u6bd4\u8bc4\u6790<\/td>\n<td>Copilot\u5728\u57fa\u7840\u7b97\u6cd5\u4efb\u52a1\u4e0a\u6709\u6548\uff0c\u4f46\u590d\u6742\u60c5\u5883\u4e0b\u9700\u8981\u4eba\u7c7b\u5224\u65ad<\/td>\n<\/tr>\n<\/table>\n<p><\/p>\n<h2>5. \u521b\u9020\u6027\u5de5\u4f5c\u4e0e\u521b\u65b0<\/h2>\n<h3>5.1 AI\u589e\u5f3a\u4e2a\u4f53\u521b\u9020\u529b\uff0c\u4f46\u51cf\u5c11\u96c6\u4f53\u591a\u6837\u6027<\/h3>\n<p>Doshi & Hauser (2024) \u5728 <em>Science Advances<\/em> \u53d1\u8868\uff1aAI\u8f85\u52a9\u4f7f\u4e2a\u4f53\u521b\u4f5c\u7684\u6545\u4e8b\u88ab\u8bc4\u4e3a\u66f4\u6709\u521b\u9020\u529b\u3001\u6587\u7b14\u66f4\u597d\u2014\u2014\u5c24\u5176\u5bf9\u521b\u9020\u529b\u8f83\u5f31\u7684\u5199\u4f5c\u8005\u3002\u4f46AI\u751f\u6210\u7684\u6545\u4e8b\u5f7c\u6b64<em>\u66f4\u52a0\u96f7\u540c<\/em>\uff0c\u964d\u4f4e\u4e86\u96c6\u4f53\u591a\u6837\u6027\u3002<\/p>\n<h3>5.2 \u7fa4\u4f53\u667a\u6167 vs. AI\u521b\u610f<\/h3>\n<p>Boussioux et al. (2024) \u5728 <em>Organization Science<\/em> \u63d0\u51fa\"\u65e0\u4eba\u7fa4\u7684\u672a\u6765\"\u6982\u5ff5\uff1a\u4eba\u7c7b\u5f15\u5bfc\u7684AI\u5408\u4f5c\u53ef\u4ee5\u589e\u5f3a\u521b\u610f\u95ee\u9898\u89e3\u51b3\uff0c\u4f46\u9700\u8981\u8c28\u614e\u8bbe\u8ba1\u534f\u4f5c\u6a21\u5f0f\u3002<\/p>\n<h3>5.3 AI\u8f85\u52a9\u4eba\u7c7b\u521b\u610f\u4efb\u52a1\u7684\u5b9e\u9a8c\u8bc1\u636e<\/h3>\n<p>\u53e6\u4e00\u9879\u7814\u7a76\uff082026, <em>Journal of Economic Behavior & Organization<\/em>\uff09\u4ee5302\u540d\u5b66\u751f\u5b8c\u62104\u9879\u65b0\u521b\u610f\u4efb\u52a1\uff0c\u53d1\u73b0\u968f\u673a\u5206\u914dAI\u8bbf\u95ee\u7684\u53c2\u4e0e\u8005\u4ea7\u51fa\u4e0e\u7eaf\u4eba\u5de5\u7ec4\u5b58\u5728\u663e\u8457\u5dee\u5f02\u3002<\/p>\n<h3>5.4 \u4eba\u7c7b-AI\u534f\u4f5c\u589e\u5f3a\u5373\u65f6\u8868\u73b0\u4f46......<\/h3>\n<p>Nature <em>Scientific Reports<\/em> (2025) \u7684\u7814\u7a76\u4e00\u81f4\u53d1\u73b0GenAI\u534f\u4f5c\u589e\u5f3a\u5373\u65f6\u4efb\u52a1\u8868\u73b0\uff0c\u4f46\u8fd9\u79cd\u589e\u5f3a\u6548\u5e94<em>\u4e0d\u4f1a\u5ef6\u7eed\u5230\u540e\u7eed\u72ec\u7acb\u4efb\u52a1\u4e2d<\/em>\u3002<\/p>\n<p><\/p>\n<h2>6. \u6700\u4f18\u4eba\u673a\u534f\u4f5c\u7b56\u7565\u7814\u7a76<\/h2>\n<h3>6.1 \u82cf\u683c\u62c9\u5e95\u5f0fAI\u5bfc\u5e08<\/h3>\n<p>\u591a\u9879\u6700\u65b0\u7814\u7a76\u4e00\u81f4\u8868\u660e\uff0c\u5c06LLM\u914d\u7f6e\u4e3a\u82cf\u683c\u62c9\u5e95\u5f0f\u5bfc\u5e08\uff08\u901a\u8fc7\u63d0\u95ee\u5f15\u5bfc\u800c\u975e\u76f4\u63a5\u7ed9\u7b54\u6848\uff09\u53ef\u4ee5\u540c\u65f6\u5b9e\u73b0\u6548\u7387\u63d0\u5347\u548c\u6280\u80fd\u4fdd\u6301\uff1a<\/p>\n<ul>\n<li><strong>SocraticAI (Sunil & Thakkar, 2024)\uff1a<\/strong>\u5c06LLM\u91cd\u6784\u4e3a\u6709\u7ea6\u675f\u7684CS\u6559\u5b66\u5bfc\u5e08\uff0c\u901a\u8fc7\u7ed3\u6784\u5316\u4e92\u52a8\u4fc3\u8fdb\u5b66\u751f\u8868\u8fbe\u63a8\u7406\u8fc7\u7a0b\u3002<\/li>\n<li><strong>STAP (2025)\uff1a<\/strong>\u82cf\u683c\u62c9\u5e95\u5f0f\u81ea\u9002\u5e94\u7f16\u7a0b\u5bfc\u5e08\uff0c\u5c06LLM\u89d2\u8272\u4ece\"\u795e\u8c15\"\u53d8\u4e3a\"\u82cf\u683c\u62c9\u5e95\u5bfc\u5e08\"\u3002<\/li>\n<li><strong>EULER (2024)\uff1a<\/strong>\u5fae\u8c03LLM\u5b9e\u73b0\u82cf\u683c\u62c9\u5e95\u5f0f\u4e92\u52a8\uff0c\u7528\u63d0\u95ee\u5f15\u5bfc\u5b66\u751f\u81ea\u5df1\u53d1\u73b0\u7b54\u6848\u3002<\/li>\n<li><strong>Khan Academy + Khanmigo\uff1a<\/strong>\u5b9a\u5236\u63d0\u793a\u4f7fAI\u4f7f\u7528\u82cf\u683c\u62c9\u5e95\u6cd5\uff0c\u51e0\u4e4e\u6bcf\u8f6e\u5bf9\u8bdd\u90fd\u63d0\u95ee\u3002<\/li>\n<li><strong>RL\u5bf9\u9f50\u6559\u5b66\u6cd5 (2025)\uff1a<\/strong>\u7528\u5f3a\u5316\u5b66\u4e60\u8bad\u7ec3LLM\u5bfc\u5e08\u4f7f\u7528\u82cf\u683c\u62c9\u5e95\u5f0f\u63d0\u95ee\u548c\u9488\u5bf9\u6027\u63d0\u793a\u3002<\/li>\n<\/ul>\n<h3>6.2 \u811a\u624b\u67b6\u5f0fAI\u8f85\u52a9<\/h3>\n<ul>\n<li><strong>DBox (2025, CHI '25)\uff1a<\/strong>\u901a\u8fc7\u5b66\u4e60\u8005-AI\u5171\u540c\u5206\u89e3\u95ee\u9898\u7684\u4ea4\u4e92\u6b65\u9aa4\u6811\uff0c\u652f\u6301\u6784\u601d\u548c\u5b9e\u73b0\u9636\u6bb5\uff0c\u540c\u65f6<em>\u57f9\u517b\u72ec\u7acb\u601d\u7ef4<\/em>\u3002<\/li>\n<li><strong>AI-based scaffolding (2025)\uff1a<\/strong>AI\u811a\u624b\u67b6\u5bf9\u5b66\u4e60\u8005\u7684\u95ee\u9898\u89e3\u51b3\u80fd\u529b\u548c\u5143\u8ba4\u77e5\u610f\u8bc6\u6709\u663e\u8457\u6b63\u5411\u5f71\u54cd\u3002<\/li>\n<li><strong>More AI Assistance Reduces Cognitive Engagement (2025)\uff1a<\/strong>AI\u8f85\u52a9\u6c34\u5e73\u8d8a\u9ad8\uff0c\u8ba4\u77e5\u53c2\u4e0e\u5ea6\u8d8a\u4f4e\u2014\u2014\u5b58\u5728\"AI\u8f85\u52a9\u56f0\u5883\"\u3002<\/li>\n<\/ul>\n<h3>6.3 \u7ed3\u5bf9\u7f16\u7a0b\u4e2d\u7684AI<\/h3>\n<p>Fast and Forgettable (2025)\uff1a\u5bf9\u6bd4AI\u8f85\u52a9\u7ed3\u5bf9\u7f16\u7a0b\u4e0e\u4f20\u7edf\u7ed3\u5bf9\u7f16\u7a0b\u7684\u65b0\u624b\u5b66\u4e60\u6548\u679c\uff0c\u53d1\u73b0AI\u8f85\u52a9\u53ef\u80fd\u66f4\u597d\u4fdd\u5b88\u5730\u4f7f\u7528\uff0c\u4e0e\u4f20\u7edf\u6a21\u5f0f\u7ed3\u5408\u3002<\/p>\n<p><\/p>\n<h2>7. \u7efc\u5408\u5206\u6790\u4e0e\u5b9e\u8df5\u5efa\u8bae<\/h2>\n<div class=\"box\">\n<strong>\u6838\u5fc3\u77db\u76fe\uff1a<\/strong>AI\u63d0\u5347\u77ed\u671f\u6548\u7387\uff0c\u4f46\u4ee5\u957f\u671f\u6280\u80fd\u9000\u5316\u4e3a\u4ee3\u4ef7\u3002\u8fd9\u4e00\u53d1\u73b0\u8d2f\u7a7f\u6240\u6709\u6587\u732e\u2014\u2014\u4ece\u5b9e\u9a8c\u5ba4RCT\u5230\u5927\u89c4\u6a21\u73b0\u573a\u5b9e\u9a8c\u3002\n<\/div>\n<h3>7.1 \u8bc1\u636e\u6c47\u603b\u8868<\/h3>\n<table>\n<tr>\n<th>\u7ef4\u5ea6<\/th>\n<th>AI\u8f85\u52a9\u66f4\u4f18<\/th>\n<th>\u4eba\u7c7b\u72ec\u7acb\u66f4\u4f18<\/th>\n<th>\u8bc1\u636e\u7b49\u7ea7<\/th>\n<\/tr>\n<tr>\n<td>\u5e38\u89c4\/\u5df2\u77e5\u590d\u6742\u95ee\u9898<\/td>\n<td>\u2705 \u6548\u7387+26~55%<\/td>\n<td>\u2014<\/td>\n<td>\u5f3a\uff083\u9879\u5927\u89c4\u6a21RCT\uff09<\/td>\n<\/tr>\n<tr>\n<td>\u524d\u6cbf\u5916\u5f00\u521b\u6027\u95ee\u9898<\/td>\n<td>\u2014<\/td>\n<td>\u2705 \u6b63\u786e\u7387+19%<\/td>\n<td>\u5f3a\uff08BCG RCT\uff09<\/td>\n<\/tr>\n<tr>\n<td>\u72ec\u7acb\u89e3\u9898\u80fd\u529b\uff08\u6280\u80fd\u8f6c\u79fb\uff09<\/td>\n<td>\u2014<\/td>\n<td>\u2705 \u6b63\u786e\u7387+16%<\/td>\n<td>\u5f3a\uff083\u9879RCT, N=1,222\uff09<\/td>\n<\/tr>\n<tr>\n<td>\u5b66\u4e60\u65b0\u6280\u80fd\u7684\u6548\u679c<\/td>\n<td>\u2014<\/td>\n<td>\u2705 \u6982\u5ff5\u7406\u89e3\u66f4\u4f18<\/td>\n<td>\u4e2d\uff08Shen & Tamkin RCT\uff09<\/td>\n<\/tr>\n<tr>\n<td>\u521b\u9020\u6027\u4efb\u52a1\uff08\u4e2a\u4f53\u5c42\u9762\uff09<\/td>\n<td>\u2705 \u521b\u610f\u8bc4\u5206+30~40%<\/td>\n<td>\u2014<\/td>\n<td>\u4e2d<\/td>\n<\/tr>\n<tr>\n<td>\u521b\u9020\u6027\u4efb\u52a1\uff08\u96c6\u4f53\u591a\u6837\u6027\uff09<\/td>\n<td>\u2014<\/td>\n<td>\u2705 \u5185\u5bb9\u591a\u6837\u6027\u66f4\u9ad8<\/td>\n<td>\u4e2d<\/td>\n<\/tr>\n<\/table>\n<h3>7.2 \u5b9e\u8df5\u7b56\u7565\u2014\u2014\"\u534a\u4eba\u9a6c\"\u6a21\u5f0f<\/h3>\n<ol>\n<li><strong>AI\u505a\u57fa\u7840\u7814\u7a76\/\u6a21\u5f0f\u8bc6\u522b\/\u6837\u677f\u4ee3\u7801\u751f\u6210\uff1a<\/strong>\u5229\u7528AI\u7684\u901f\u5ea6\u548c\u5e7f\u5ea6\u4f18\u52bf\u3002<\/li>\n<li><strong>\u4eba\u505a\u67b6\u6784\u5224\u65ad\/\u521b\u65b0\/\u4e0a\u4e0b\u6587\u7406\u89e3\/\u5f02\u5e38\u5904\u7406\uff1a<\/strong>\u4fdd\u7559\u4eba\u7c7b\u72ec\u6709\u7684\u5224\u65ad\u529b\u3002<\/li>\n<li><strong>\u82cf\u683c\u62c9\u5e95\u5f0fPrompt\uff1a<\/strong>\u660e\u786e\u6307\u793aAI\"\u901a\u8fc7\u63d0\u95ee\u5f15\u5bfc\uff0c\u4e0d\u8981\u76f4\u63a5\u7ed9\u7b54\u6848\"\u3002<\/li>\n<li><strong>\u81ea\u6211\u8bc4\u4f30\u673a\u5236\uff1a<\/strong>\u5b9a\u671f\u5728\u65e0AI\u73af\u5883\u4e0b\u6d4b\u8bd5\u72ec\u7acb\u80fd\u529b\uff0c\u9632\u6b62\u6280\u80fd\u9000\u5316\u3002<\/li>\n<li><strong>\u5bf9\u65b0\u624b\u7684\u5173\u952e\u5efa\u8bae\uff1a<\/strong>\u5728\u65f6\u95f4\u538b\u529b\u4e0b\uff0c\u7528AI\u83b7\u53d6\u77e5\u8bc6\u6846\u67b6\u548c\u6587\u732e\u7d22\u5f15\uff0c\u4f46\u6838\u5fc3\u5b9e\u73b0\u548c\u67b6\u6784\u51b3\u7b56\u7531\u81ea\u5df1\u5b8c\u6210\u3002<\/li>\n<\/ol>\n<p><\/p>\n<div class=\"ref-section\">\n<h2>\u53c2\u8003\u6587\u732e<\/h2>\n<p class=\"ref\">Boussioux, L., Lane, J. N., Zhang, M., Vladimirsky, V., & Lakhani, K. R. (2024). The crowdless future? Generative AI and creative problem-solving. <em>Organization Science<\/em>, <em>35<\/em>(5), 1589\u20131607. https:\/\/doi.org\/10.1287\/orsc.2023.18430<\/p>\n<p class=\"ref\">Brynjolfsson, E., Li, D., & Raymond, L. R. (2025). Generative AI at work. <em>The Quarterly Journal of Economics<\/em>, <em>140<\/em>(2), 889\u2013942. https:\/\/doi.org\/10.1093\/qje\/qjae031<\/p>\n<p class=\"ref\">Cui, Z., Demirer, M., Jaffe, S., Musolff, L., Peng, S., & Salz, T. (2025). The effects of generative AI on high-skilled work: Evidence from three field experiments. <em>Management Science<\/em>. Advance online publication. https:\/\/doi.org\/10.1287\/mnsc.2025.00535<\/p>\n<p class=\"ref\">Dell'Acqua, F., McFowland, E., III, Mollick, E. R., Lifshitz-Assaf, H., Kellogg, K., Rajendran, S., Krayer, L., Candelon, F., & Lakhani, K. R. (2025). Navigating the jagged technological frontier: Field experimental evidence of the effects of AI on knowledge worker productivity and quality. <em>Organization Science<\/em>. Advance online publication. https:\/\/doi.org\/10.1287\/orsc.2025.21838<\/p>\n<p class=\"ref\">Doshi, A. R., & Hauser, O. (2024). Generative AI enhances individual creativity but reduces the collective diversity of novel content. <em>Science Advances<\/em>, <em>10<\/em>(29), eadn5290. https:\/\/doi.org\/10.1126\/sciadv.adn5290<\/p>\n<p class=\"ref\">Hohenstein, J., Kizilcec, R. F., DiGiacomo, D., Aghajari, Z., Mubarrat, S., Biehl, M., & Jung, M. F. (2024). Does using artificial intelligence assistance accelerate skill decay and hinder skill development without performers' awareness? <em>Cognitive Research: Principles and Implications<\/em>, <em>9<\/em>, Article 46. https:\/\/doi.org\/10.1186\/s41235-024-00572-8<\/p>\n<p class=\"ref\">Lee, H. P., Sarkar, A., Tankelevitch, L., Drosos, I., Rintel, S., Banks, R., & Wilson, N. (2025). The impact of generative AI on critical thinking: Self-reported reductions in cognitive effort and confidence effects from a survey of knowledge workers. In <em>Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems<\/em> (CHI '25). ACM. https:\/\/doi.org\/10.1145\/3706598.3713778<\/p>\n<p class=\"ref\">Liu, G., Christian, B., Dumbalska, T., Bakker, M. A., & Dubey, R. (2025). AI assistance reduces persistence and hurts independent performance. <em>arXiv preprint arXiv:2604.04721<\/em>. https:\/\/arxiv.org\/abs\/2604.04721<\/p>\n<p class=\"ref\">Noy, S., & Zhang, W. (2023). Experimental evidence on the productivity effects of generative artificial intelligence. <em>Science<\/em>, <em>381<\/em>(6654), 187\u2013192. https:\/\/doi.org\/10.1126\/science.adh2586<\/p>\n<p class=\"ref\">Peng, S., Kalliamvakou, E., Cihon, P., & Demirer, M. (2023). The impact of AI on developer productivity: Evidence from GitHub Copilot. <em>arXiv preprint arXiv:2302.06590<\/em>. https:\/\/arxiv.org\/abs\/2302.06590<\/p>\n<p class=\"ref\">Shen, J. H., & Tamkin, A. (2026). How AI impacts skill formation: Evidence from software development learning. <em>arXiv preprint arXiv:2601.20245<\/em>. https:\/\/arxiv.org\/abs\/2601.20245<\/p>\n<p class=\"ref\">Vaccaro, M., Almaatouq, A., & Malone, T. W. (2024). When combinations of humans and AI are useful: A systematic review and meta-analysis. <em>Nature Human Behaviour<\/em>, <em>8<\/em>, 2295\u20132308. https:\/\/doi.org\/10.1038\/s41562-024-02024-1<\/p>\n<h3 style=\"margin-top: 24px;\">\u8865\u5145\u6587\u732e\u2014\u2014\u6700\u4f18\u534f\u4f5c\u7b56\u7565<\/h3>\n<p class=\"ref\">Chowdhury, A., Roy, S., & Radev, D. (2024). MWPTutor: LLM-powered tutoring with hierarchical guardrails. In <em>Proceedings of the AAAI Conference on Artificial Intelligence<\/em>.<\/p>\n<p class=\"ref\">Dhillon, H., Gu, Z., & Shneiderman, B. (2024). More AI assistance reduces cognitive engagement: Examining the AI assistance dilemma in AI-supported note-taking. <em>arXiv preprint arXiv:2509.03392<\/em>. https:\/\/arxiv.org\/abs\/2509.03392<\/p>\n<p class=\"ref\">Gatti, G., & colleagues. (2024). EULER: Fine tuning a large language model for Socratic interactions. In <em>Proceedings of AIxEDU 2024<\/em> (CEUR Workshop Proceedings, Vol. 3879).<\/p>\n<p class=\"ref\">Kazemitabaar, M., Chow, J., Ma, C. K. T., Ericson, B. J., Weintrop, D., & Grossman, T. (2025). Fast and forgettable: A controlled study of novices' performance, learning, workload, and emotion in AI-assisted and human pair programming paradigms. <em>arXiv preprint arXiv:2604.18538<\/em>. https:\/\/arxiv.org\/abs\/2604.18538<\/p>\n<p class=\"ref\">Liu, J., Huang, Z., & colleagues. (2024). SocraticLM: Exploring Socratic personalized teaching with large language models. In <em>Advances in Neural Information Processing Systems<\/em> (NeurIPS 2024).<\/p>\n<p class=\"ref\">Sunil, S., & Thakkar, P. (2024). SocraticAI: Transforming LLMs into guided CS tutors through scaffolded interaction. <em>arXiv preprint arXiv:2512.03501<\/em>. https:\/\/arxiv.org\/abs\/2512.03501<\/p>\n<p class=\"ref\">Wang, Z., Liu, Y., & Chen, X. (2025). STAP: A Socratic tutor for adaptive programming with pedagogical scaffolding. In <em>Proceedings of the 56th ACM Technical Symposium on Computer Science Education<\/em> (SIGCSE TS 2025). ACM. https:\/\/doi.org\/10.1145\/3775073.3775165<\/p>\n<p class=\"ref\">Yang, K., Zhang, T., & Li, S. (2025). From problem-solving to teaching problem-solving: Aligning LLMs with pedagogy using reinforcement learning. <em>arXiv preprint arXiv:2505.15607<\/em>. https:\/\/arxiv.org\/abs\/2505.15607<\/p>\n<p class=\"ref\">Zhang, L., Ma, N., & Zhao, G. (2024). Generative AI in education: From foundational insights to the Socratic playground for learning. <em>arXiv preprint arXiv:2501.06682<\/em>. https:\/\/arxiv.org\/abs\/2501.06682<\/p>\n<p class=\"ref\">Zhao, Y., Wu, R., & Peng, H. (2025). DBox: Scaffolding algorithmic programming learning through learner-LLM co-decomposition. In <em>Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems<\/em> (CHI '25). ACM. https:\/\/doi.org\/10.1145\/3706598.3713748<\/p>\n<p class=\"ref\">(2025). Human-generative AI collaboration enhances task performance but does not persist in subsequent tasks. <em>Scientific Reports<\/em>. https:\/\/doi.org\/10.1038\/s41598-025-98385-2<\/p>\n<p class=\"ref\">(2025). AI-assisted vs human-only evidence review: Results from a comparative study. UK Government Behavioural Insights Team. https:\/\/www.gov.uk\/government\/publications\/ai-assisted-vs-human-only-evidence-review<\/p>\n<p class=\"ref\">(2026). Generative AI adoption in human creative tasks: Experimental evidence. <em>Journal of Economic Behavior & Organization<\/em>. https:\/\/doi.org\/10.1016\/j.jebo.2026.01.002<\/p>\n<p class=\"ref\">(2025). Can pedagogical agent-based scaffolding boost information problem solving? <em>Education and Information Technologies<\/em>. https:\/\/doi.org\/10.1007\/s10639-025-13784-2<\/p>\n<p class=\"ref\">(2025). The effect of AI-based scaffolding on problem solving and metacognitive awareness in learners. <em>ResearchGate<\/em>. https:\/\/doi.org\/10.13140\/RG.2.2.33921.47201<\/p>\n<\/div>\n<p style=\"text-align:center;margin-top:40px;color:#999;font-size:0.85em;\">\n\u8986\u76d6\u6587\u732e\u6570\uff1a27\u7bc7 \u00b7 \u7c7b\u578b\uff1a\u5143\u5206\u6790\/\u7cfb\u7edf\u7efc\u8ff0\/RCT\/\u73b0\u573a\u5b9e\u9a8c\/\u7406\u8bba\u5206\u6790 \u00b7 \u8de8\u5ea6\uff1a2023\u20132026\n<\/p>\n<\/div>\n<p><\/body><br \/>\n<\/html><\/p>\n","protected":false},"excerpt":{"rendered":"<p>AI\u8f85\u52a9 vs \u4eba\u7c7b\u72ec\u7acb\u89e3\u51b3\u95ee\u9898\u6548\u7387\u5bf9\u6bd4\u2014\u2014\u6587\u732e\u7efc\u8ff0 AI\u8f85\u52a9\u4e0e\u4eba\u7c7b\u72ec\u7acb\u89e3\u51b3\u95ee\u9898\u6548\u7387\u5bf9\u6bd4 \u2014\u2014\u805a\u7126\u590d\u6742\/\u5f00\u521b\u6027\u95ee\u9898\u7684\u6587\u732e\u7efc\u8ff0 \u6838\u5fc3\u95ee\u9898\uff1a\u5bf9\u4e8e\u56f0\u96be\u7684\u5f00\u521b\u6027\u95ee\u9898\uff08\u5982\u5b9a\u5236\u4e0b\u8f7d\u534f\u8bae\u3001\u5e26\u5bbd\u901f\u7387\u9650\u5236\u7b49\uff09\uff0c\u4eba\u7c7b\u5229\u7528AI\u8f85\u52a9\u4e0e\u901a\u8fc7","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"emotion":"","emotion_color":"","title_style":"","license":"","footnotes":""},"categories":[1],"tags":[],"class_list":["post-116","post","type-post","status-publish","format-standard","hentry","category-uncategorized"],"_links":{"self":[{"href":"https:\/\/www.tom-thu.cn\/index.php?rest_route=\/wp\/v2\/posts\/116","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.tom-thu.cn\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.tom-thu.cn\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.tom-thu.cn\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.tom-thu.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=116"}],"version-history":[{"count":1,"href":"https:\/\/www.tom-thu.cn\/index.php?rest_route=\/wp\/v2\/posts\/116\/revisions"}],"predecessor-version":[{"id":117,"href":"https:\/\/www.tom-thu.cn\/index.php?rest_route=\/wp\/v2\/posts\/116\/revisions\/117"}],"wp:attachment":[{"href":"https:\/\/www.tom-thu.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=116"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.tom-thu.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=116"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.tom-thu.cn\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=116"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}

{"id":116,"date":"2026-05-10T01:33:18","date_gmt":"2026-05-09T17:33:18","guid":{"rendered":"https:\/\/www.tom-thu.cn\/?p=116"},"modified":"2026-05-10T01:33:19","modified_gmt":"2026-05-09T17:33:19","slug":"%e3%80%90%e8%a7%82%e5%af%9f%e3%80%91ai%e7%9c%9f%e7%9a%84%e6%8f%90%e9%ab%98%e4%ba%86%e8%a7%a3%e5%86%b3%e9%97%ae%e9%a2%98%e7%9a%84%e6%95%88%e7%8e%87%e4%ba%86%e5%90%97%ef%bc%9f%ef%bc%88%e4%b8%8a","status":"publish","type":"post","link":"https:\/\/www.tom-thu.cn\/?p=116","title":{"rendered":"\u3010\u89c2\u5bdf\u3011AI\u771f\u7684\u63d0\u9ad8\u4e86\u89e3\u51b3\u95ee\u9898\u7684\u6548\u7387\u4e86\u5417\uff1f\uff08\u4e0a-\u6587\u732e\u7efc\u8ff0\uff09"},"content":{"rendered":"

\n
\n
\n
\n
\nAI\u8f85\u52a9 vs \u4eba\u7c7b\u72ec\u7acb\u89e3\u51b3\u95ee\u9898\u6548\u7387\u5bf9\u6bd4\u2014\u2014\u6587\u732e\u7efc\u8ff0<\/title><\/p>\n<style>\n * { box-sizing: border-box; margin: 0; padding: 0; }\n body { font-family: 'Times New Roman', Georgia, serif; line-height: 1.8; color: #222; background: #fafafa; }\n .container { max-width: 900px; margin: 0 auto; padding: 40px 30px; background: #fff; }\n h1 { font-size: 1.8em; text-align: center; margin-bottom: 8px; color: #1a1a1a; }\n h2 { font-size: 1.3em; margin: 40px 0 16px; border-bottom: 2px solid #333; padding-bottom: 6px; color: #1a1a1a; }\n h3 { font-size: 1.1em; margin: 24px 0 10px; color: #333; }\n p { margin: 12px 0; text-align: justify; }\n ul, ol { margin: 10px 0 10px 30px; }\n li { margin: 6px 0; }\n .highlight { background: #fff3cd; border-left: 4px solid #ffc107; padding: 12px 16px; margin: 16px 0; }\n .box { background: #f0f4ff; border-left: 4px solid #4a90d9; padding: 12px 16px; margin: 16px 0; }\n .meta { text-align: center; color: #666; font-size: 0.9em; margin-bottom: 32px; }\n .ref { padding-left: 30px; text-indent: -30px; margin: 8px 0; font-size: 0.95em; }\n .ref-section { margin-top: 40px; border-top: 1px solid #ccc; padding-top: 20px; }\n table { width: 100%; border-collapse: collapse; margin: 16px 0; font-size: 0.9em; }\n th, td { border: 1px solid #ddd; padding: 10px 12px; text-align: left; }\n th { background: #4a90d9; color: #fff; font-weight: bold; }\n tr:nth-child(even) { background: #f8f8f8; }\n .tag { display: inline-block; padding: 2px 8px; border-radius: 3px; font-size: 0.8em; font-weight: bold; margin-right: 4px; }\n .tag-pos { background: #d4edda; color: #155724; }\n .tag-neg { background: #f8d7da; color: #721c24; }\n .tag-mix { background: #fff3cd; color: #856404; }\n @media (max-width: 600px) {\n .container { padding: 20px 15px; }\n h1 { font-size: 1.4em; }\n }\n<\/style>\n<p><\/head><br \/>\n<body><\/p>\n<div class=\"container\">\n<h1>AI\u8f85\u52a9\u4e0e\u4eba\u7c7b\u72ec\u7acb\u89e3\u51b3\u95ee\u9898\u6548\u7387\u5bf9\u6bd4<\/h1>\n<h1 style=\"font-size:1.2em;font-weight:normal;color:#555;margin-bottom:20px;\">\u2014\u2014\u805a\u7126\u590d\u6742\/\u5f00\u521b\u6027\u95ee\u9898\u7684\u6587\u732e\u7efc\u8ff0<\/h1>\n<div class=\"highlight\">\n<strong>\u6838\u5fc3\u95ee\u9898\uff1a<\/strong>\u5bf9\u4e8e\u56f0\u96be\u7684\u5f00\u521b\u6027\u95ee\u9898\uff08\u5982\u5b9a\u5236\u4e0b\u8f7d\u534f\u8bae\u3001\u5e26\u5bbd\u901f\u7387\u9650\u5236\u7b49\uff09\uff0c\u4eba\u7c7b\u5229\u7528AI\u8f85\u52a9\u4e0e\u901a\u8fc7\u72ec\u7acb\u6587\u732e\u7814\u7a76\u4e24\u79cd\u65b9\u5f0f\uff0c\u89e3\u51b3\u95ee\u9898\u7684\u6548\u7387\u662f\u5426\u5b58\u5728\u663e\u8457\u5dee\u5f02\uff1f\n<\/div>\n<p><\/p>\n<h2>1. \u5143\u5206\u6790\uff1a\u4eba+AI\u4f55\u65f6\u6709\u7528\uff1f<\/h2>\n<p>\u8fc4\u4eca\u4e3a\u6b62\u6700\u5168\u9762\u7684\u4eba\u673a\u534f\u4f5c\u5143\u5206\u6790\u7531MIT Sloan\u56e2\u961f\u5b8c\u6210\uff0c\u8986\u76d6106\u9879\u5b9e\u9a8c\uff08370\u4e2a\u6548\u5e94\u91cf\uff09\uff0c\u53d1\u8868\u4e8e <em>Nature Human Behaviour<\/em>\uff08Vaccaro et al., 2024\uff09\u3002<\/p>\n<ul>\n<li><strong>\u603b\u4f53\u7ed3\u8bba\uff1a<\/strong>\u4eba+AI\u7ec4\u5408<em>\u5e73\u5747<\/em>\u8868\u73b0\u4f18\u4e8e\u7eaf\u4eba\u5de5\uff0c\u4f46\u4e0d\u5982\u7eafAI\u7cfb\u7edf\u7684\u6700\u4f73\u8868\u73b0\u3002\u672a\u53d1\u73b0\"\u4eba\u673a\u534f\u540c\u6548\u5e94\"\uff08\u5373\u7ec4\u5408\u8868\u73b0\u8d85\u8d8a\u5404\u81ea\u6700\u4f18\u503c\uff09\u3002<\/li>\n<li><strong>\u51b3\u7b56\u7c7b\u4efb\u52a1\uff1a<\/strong>\uff08\u6df1\u5ea6\u4f2a\u9020\u68c0\u6d4b\u3001\u533b\u7597\u8bca\u65ad\u3001\u9700\u6c42\u9884\u6d4b\u7b49\uff09\u4eba+AI\u7ec4\u5408\u5e38<em>\u4e0d\u5982<\/em>\u7eafAI\u3002<\/li>\n<li><strong>\u521b\u9020\u6027\u4efb\u52a1\uff1a<\/strong>\uff08\u6587\u672c\u603b\u7ed3\u3001\u95ee\u7b54\u804a\u5929\u3001\u56fe\u50cf\u751f\u6210\u3001\u65b0\u5185\u5bb9\u521b\u4f5c\u7b49\uff09\u4eba+AI\u7ec4\u5408\u5e38<em>\u8d85\u8d8a<\/em>\u5404\u81ea\u6700\u4f73\u6c34\u5e73\u3002<\/li>\n<li><strong>\u6838\u5fc3\u6d1e\u5bdf\uff1a<\/strong>\"\u6709\u6548\u6027\u4e0d\u5728\u4e8e\u4efb\u4f55\u4e00\u65b9\u7684\u57fa\u7ebf\u8868\u73b0\uff0c\u800c\u5728\u4e8e\u4e8c\u8005\u5982\u4f55\u534f\u4f5c\u548c\u4e92\u8865\u3002\"<\/li>\n<\/ul>\n<p><\/p>\n<h2>2. AI\u8f85\u52a9\u635f\u5bb3\u72ec\u7acb\u89e3\u9898\u80fd\u529b\u2014\u2014\u56e0\u679c\u8bc1\u636e<\/h2>\n<h3>2.1 \u4ec510\u5206\u949fAI\u8f85\u52a9\u5373\u9020\u6210\u663e\u8457\u635f\u4f24<\/h3>\n<p>Liu et al. (2025) \u901a\u8fc73\u9879RCT\uff08N=1,222\uff09\uff0c\u63d0\u4f9b\u4e86\u4e00\u5957\u5b8c\u6574\u7684\u56e0\u679c\u8bc1\u636e\u94fe\uff1a<\/p>\n<ul>\n<li><strong>\u5b9e\u9a8c1\uff08N=354\uff09\uff1a<\/strong>\u6570\u5b66\u89e3\u9898\u4efb\u52a1\u3002AI\u8f85\u52a9\u7ec4\u5728\u64a4\u6389AI\u540e\uff0c\u72ec\u7acb\u89e3\u9898\u6b63\u786e\u7387\u4ece73%\u964d\u81f357%\uff08Cohen's <em>d<\/em> = \u22120.42\uff09\uff0c\u653e\u5f03\u7387\u4ece11%\u5347\u81f320%\u3002<\/li>\n<li><strong>\u5b9e\u9a8c2\uff08N=667\uff09\uff1a<\/strong>\u590d\u73b0\u5e76\u6392\u9664\u6df7\u6dc6\u53d8\u91cf\uff08\u589e\u52a0\u524d\u6d4b\u3001\u7edf\u4e00\u754c\u9762\uff09\uff0c\u6548\u5e94\u590d\u73b0\uff08\u6b63\u786e\u738771% vs 77%\uff09\u3002<\/li>\n<li><strong>\u5b9e\u9a8c3\uff08N=201\uff09\uff1a<\/strong>\u9605\u8bfb\u7406\u89e3\u4efb\u52a1\uff08SAT\u98ce\u683c\uff09\u3002\u540c\u7b49\u6548\u5e94\u590d\u73b0\uff08\u6b63\u786e\u738776% vs 89%\uff0c<em>d<\/em> = \u22120.42\uff09\uff0c\u8bc1\u660e\u6548\u5e94\u5177\u6709\u8de8\u9886\u57df\u6cdb\u5316\u6027\u3002<\/li>\n<li><strong>\u5173\u952e\u53d1\u73b0\uff1a<\/strong>61%\u7684\u7528\u6237\u76f4\u63a5\u5411AI\u7d22\u8981\u7b54\u6848\uff0c\u8fd9\u90e8\u5206\u4eba\u53d7\u635f\u6700\u4e25\u91cd\uff1b<em>\u7528AI\u8981\u63d0\u793a\/\u6f84\u6e05\u7684\u7528\u6237\u4e0e\u5bf9\u7167\u7ec4\u65e0\u663e\u8457\u5dee\u5f02<\/em>\u3002<\/li>\n<\/ul>\n<h3>2.2 AI\u52a0\u901f\u6280\u80fd\u9000\u5316<\/h3>\n<p>Hohenstein et al. (2024) \u4ece\u7406\u8bba\u89d2\u5ea6\u5206\u6790\uff1aAI\u52a9\u624b\u53ef\u80fd\u52a0\u901f\u4e13\u5bb6\u7684\u6280\u80fd\u9000\u5316\uff0c\u963b\u788d\u65b0\u624b\u7684\u6280\u80fd\u4e60\u5f97\uff0c\u4e14\u7528\u6237\u5f80\u5f80\u610f\u8bc6\u4e0d\u5230\u8fd9\u4e9b\u8d1f\u9762\u5f71\u54cd\u3002<\/p>\n<h3>2.3 \u6280\u80fd\u5f62\u6210\u7684\u5b9e\u9a8c\u8bc1\u636e<\/h3>\n<p>Shen & Tamkin (2026) \u752852\u540d\u4e13\u4e1a\u5f00\u53d1\u8005\u5b66\u4e60\u65b0\u5f02\u6b65\u7f16\u7a0b\u5e93\uff08Python Trio\uff09\u7684RCT\u8bc1\u660e\uff1aAI\u7ec4\u7684\u6982\u5ff5\u7406\u89e3\u3001\u4ee3\u7801\u9605\u8bfb\u548c\u8c03\u8bd5\u80fd\u529b\u5747\u663e\u8457\u53d7\u635f\uff0c\u4e14<em>\u5e73\u5747\u6548\u7387\u5e76\u65e0\u663e\u8457\u63d0\u5347<\/em>\u3002\u5b8c\u5168\u59d4\u6258AI\u7f16\u7801\u7684\u7528\u6237\u6709\u8f7b\u5fae\u751f\u4ea7\u529b\u6539\u5584\uff0c\u4f46\u4ee5\u5b8c\u5168\u672a\u5b66\u4f1a\u8be5\u5e93\u4e3a\u4ee3\u4ef7\u3002\u7814\u7a76\u8005\u8bc6\u522b\u51fa<strong>6\u79cdAI\u4ea4\u4e92\u6a21\u5f0f\uff0c\u5176\u4e2d3\u79cd\u4fdd\u6301\u8ba4\u77e5\u53c2\u4e0e\u5ea6\u7684\u6a21\u5f0f\u4fdd\u7559\u4e86\u5b66\u4e60\u6548\u679c<\/strong>\u3002<\/p>\n<h3>2.4 AI\u8f85\u52a9\u5bf9\u6279\u5224\u6027\u601d\u7ef4\u7684\u5f71\u54cd<\/h3>\n<p>Lee et al. (2025) \u5bf9\u77e5\u8bc6\u578b\u5de5\u4f5c\u8005\u7684\u8c03\u67e5\u53d1\u73b0\uff1aGenAI\u5e2e\u52a9\u77e5\u8bc6\u578b\u5de5\u4f5c\u8005\u642d\u5efa\u590d\u6742\u4efb\u52a1\u6846\u67b6\u5e76\u81ea\u52a8\u5316\u5de5\u4ef6\u521b\u5efa\uff0c\u4f46\u4e5f\u5bfc\u81f4\u81ea\u6211\u62a5\u544a\u7684<em>\u8ba4\u77e5\u52aa\u529b\u51cf\u5c11<\/em>\u548c<em>\u4fe1\u5fc3\u6548\u5e94\u53d8\u5316<\/em>\u3002<\/p>\n<p><\/p>\n<h2>3. \u5927\u89c4\u6a21\u73b0\u573a\u5b9e\u9a8c<\/h2>\n<h3>3.1 \"\u952f\u9f7f\u72b6\u6280\u672f\u524d\u6cbf\"\u2014\u2014BCG\/HBS\u7814\u7a76<\/h3>\n<p>Dell'Acqua et al. (2025) \u5728Boston Consulting Group\u5bf9758\u540d\u987e\u95ee\u8fdb\u884cRCT\uff0c\u53d1\u8868\u4e8e <em>Organization Science<\/em>\uff1a<\/p>\n<table>\n<tr>\n<th>\u4efb\u52a1\u7c7b\u578b<\/th>\n<th>AI\u5bf9\u6027\u80fd\u7684\u5f71\u54cd<\/th>\n<th>\u6548\u679c\u91cf<\/th>\n<\/tr>\n<tr>\n<td>\u524d\u6cbf\u5185\u4efb\u52a1\uff08\u521b\u610f\u3001\u5206\u6790\u3001\u5199\u4f5c\u3001\u8bf4\u670d\uff09<\/td>\n<td>\u663e\u8457\u63d0\u5347<span class=\"tag tag-pos\">\u2191<\/span><\/td>\n<td>\u901f\u5ea6+25%\uff0c\u8d28\u91cf+40%\uff0c\u5b8c\u6210\u7387+12%<\/td>\n<\/tr>\n<tr>\n<td>\u524d\u6cbf\u5916\u4efb\u52a1\uff08\u590d\u6742\u6218\u7565\u51b3\u7b56\uff09<\/td>\n<td>\u663e\u8457\u964d\u4f4e<span class=\"tag tag-neg\">\u2193<\/span><\/td>\n<td>\u6b63\u786e\u65b9\u6848\u4ea7\u51fa\u7387<strong>\u964d\u4f4e19%<\/strong><\/td>\n<\/tr>\n<\/table>\n<p>\u7814\u7a76\u8bc6\u522b\u51fa\u4e24\u79cd\u6210\u529f\u4f7f\u7528\u6a21\u5f0f\uff1a<strong>\"\u534a\u4eba\u9a6c\"\uff08Centaurs\uff09<\/strong>\u2014\u2014\u4eba\u673a\u5206\u5de5\u534f\u4f5c\uff1b<strong>\"\u8d5b\u535a\u683c\"\uff08Cyborgs\uff09<\/strong>\u2014\u2014\u6df1\u5ea6\u878d\u5408\u5de5\u4f5c\u6d41\u3002<\/p>\n<h3>3.2 \u8f6f\u4ef6\u5de5\u7a0b\u5927\u89c4\u6a21RCT<\/h3>\n<p>Cui et al. (2025) \u5728Microsoft\u3001Accenture\u3001Fortune 100\u4e09\u5bb6\u4f01\u4e1a\u5bf94,867\u540d\u5f00\u53d1\u8005\u7684RCT\uff08\u53d1\u8868\u4e8e <em>Management Science<\/em>\uff09\uff1aCopilot\u7ec4\u4efb\u52a1\u5b8c\u6210\u91cf<strong>+26%<\/strong>\uff0c\u4f46\u63d0\u5347<em>\u96c6\u4e2d\u5728\u5e38\u89c4\u7f16\u7801\u4efb\u52a1<\/em>\uff0c\u5bf9\u8c03\u8bd5\u3001\u67b6\u6784\u51b3\u7b56\u7b49\u590d\u6742\u4efb\u52a1\u65e0\u663e\u8457\u5e2e\u52a9\u3002<\/p>\n<h3>3.3 \u547c\u53eb\u4e2d\u5fc3\u73b0\u573a\u5b9e\u9a8c<\/h3>\n<p>Brynjolfsson et al. (2025) \u5728 <em>Quarterly Journal of Economics<\/em> \u53d1\u8868\uff1aAI\u8f85\u52a9\u4f7f\u5ba2\u670d\u4eba\u5458\u6bcf\u5c0f\u65f6\u89e3\u51b3\u95ee\u9898\u6570\u91cf<strong>+15%<\/strong>\uff0c\u6548\u679c\u5bf9\u65b0\u5458\u5de5\u66f4\u663e\u8457\uff08+34%\uff09\uff0c\u5bf9\u9ad8\u6280\u80fd\u5458\u5de5\u5f71\u54cd\u6709\u9650\u3002<\/p>\n<h3>3.4 \u4e13\u4e1a\u5199\u4f5c\u4efb\u52a1<\/h3>\n<p>Noy & Zhang (2023) \u5728 <em>Science<\/em> \u53d1\u8868\uff1a\u4e2d\u7b49\u6c34\u5e73\u4e13\u4e1a\u5199\u4f5c\u4efb\u52a1\u4e2d\uff0cChatGPT\u663e\u8457\u63d0\u5347\u4e86\u751f\u4ea7\u529b\uff08\u4efb\u52a1\u5b8c\u6210\u65f6\u95f4\u51cf\u5c11\uff09\u548c\u8d28\u91cf\u3002<\/p>\n<p><\/p>\n<h2>4. AI\u7f16\u7a0b\u52a9\u624b\u4e13\u9879\u7814\u7a76<\/h2>\n<p>GitHub Copilot\u76f8\u5173\u7814\u7a76\u6784\u6210\u4e86\u5355\u4e00\u5de5\u5177\u7684\u6700\u5927\u8bc1\u636e\u96c6\uff1a<\/p>\n<table>\n<tr>\n<th>\u7814\u7a76<\/th>\n<th>\u8bbe\u8ba1<\/th>\n<th>\u5173\u952e\u53d1\u73b0<\/th>\n<\/tr>\n<tr>\n<td>Peng et al. (2023)<\/td>\n<td>\u5bf9\u7167\u5b9e\u9a8c\uff08HTTP\u670d\u52a1\u5668\u4efb\u52a1\uff09<\/td>\n<td>Copilot\u7ec4\u4efb\u52a1\u5b8c\u6210\u65f6\u95f4<strong>\u51cf\u5c1155.8%<\/strong><\/td>\n<\/tr>\n<tr>\n<td>Cui et al. (2025)<\/td>\n<td>3\u9879RCT\u51714,867\u4eba<\/td>\n<td>\u751f\u4ea7\u529b+26%\uff1b\u6548\u679c\u96c6\u4e2d\u4e8e\u5e38\u89c4\u4efb\u52a1<\/td>\n<\/tr>\n<tr>\n<td>Developer Productivity With\/Without Copilot (2025)<\/td>\n<td>\u7eb5\u5411\u6df7\u5408\u65b9\u6cd5\u7814\u7a76<\/td>\n<td>\u8c03\u78142,631\u540d\u5f00\u53d1\u8005\uff0c\u611f\u77e5\u751f\u4ea7\u529b\u4e0e\u5b9e\u9645\u4f7f\u7528\u6b63\u76f8\u5173<\/td>\n<\/tr>\n<tr>\n<td>GitHub Copilot: Asset or Liability? (2023)<\/td>\n<td>\u5bf9\u6bd4\u8bc4\u6790<\/td>\n<td>Copilot\u5728\u57fa\u7840\u7b97\u6cd5\u4efb\u52a1\u4e0a\u6709\u6548\uff0c\u4f46\u590d\u6742\u60c5\u5883\u4e0b\u9700\u8981\u4eba\u7c7b\u5224\u65ad<\/td>\n<\/tr>\n<\/table>\n<p><\/p>\n<h2>5. \u521b\u9020\u6027\u5de5\u4f5c\u4e0e\u521b\u65b0<\/h2>\n<h3>5.1 AI\u589e\u5f3a\u4e2a\u4f53\u521b\u9020\u529b\uff0c\u4f46\u51cf\u5c11\u96c6\u4f53\u591a\u6837\u6027<\/h3>\n<p>Doshi & Hauser (2024) \u5728 <em>Science Advances<\/em> \u53d1\u8868\uff1aAI\u8f85\u52a9\u4f7f\u4e2a\u4f53\u521b\u4f5c\u7684\u6545\u4e8b\u88ab\u8bc4\u4e3a\u66f4\u6709\u521b\u9020\u529b\u3001\u6587\u7b14\u66f4\u597d\u2014\u2014\u5c24\u5176\u5bf9\u521b\u9020\u529b\u8f83\u5f31\u7684\u5199\u4f5c\u8005\u3002\u4f46AI\u751f\u6210\u7684\u6545\u4e8b\u5f7c\u6b64<em>\u66f4\u52a0\u96f7\u540c<\/em>\uff0c\u964d\u4f4e\u4e86\u96c6\u4f53\u591a\u6837\u6027\u3002<\/p>\n<h3>5.2 \u7fa4\u4f53\u667a\u6167 vs. AI\u521b\u610f<\/h3>\n<p>Boussioux et al. (2024) \u5728 <em>Organization Science<\/em> \u63d0\u51fa\"\u65e0\u4eba\u7fa4\u7684\u672a\u6765\"\u6982\u5ff5\uff1a\u4eba\u7c7b\u5f15\u5bfc\u7684AI\u5408\u4f5c\u53ef\u4ee5\u589e\u5f3a\u521b\u610f\u95ee\u9898\u89e3\u51b3\uff0c\u4f46\u9700\u8981\u8c28\u614e\u8bbe\u8ba1\u534f\u4f5c\u6a21\u5f0f\u3002<\/p>\n<h3>5.3 AI\u8f85\u52a9\u4eba\u7c7b\u521b\u610f\u4efb\u52a1\u7684\u5b9e\u9a8c\u8bc1\u636e<\/h3>\n<p>\u53e6\u4e00\u9879\u7814\u7a76\uff082026, <em>Journal of Economic Behavior & Organization<\/em>\uff09\u4ee5302\u540d\u5b66\u751f\u5b8c\u62104\u9879\u65b0\u521b\u610f\u4efb\u52a1\uff0c\u53d1\u73b0\u968f\u673a\u5206\u914dAI\u8bbf\u95ee\u7684\u53c2\u4e0e\u8005\u4ea7\u51fa\u4e0e\u7eaf\u4eba\u5de5\u7ec4\u5b58\u5728\u663e\u8457\u5dee\u5f02\u3002<\/p>\n<h3>5.4 \u4eba\u7c7b-AI\u534f\u4f5c\u589e\u5f3a\u5373\u65f6\u8868\u73b0\u4f46......<\/h3>\n<p>Nature <em>Scientific Reports<\/em> (2025) \u7684\u7814\u7a76\u4e00\u81f4\u53d1\u73b0GenAI\u534f\u4f5c\u589e\u5f3a\u5373\u65f6\u4efb\u52a1\u8868\u73b0\uff0c\u4f46\u8fd9\u79cd\u589e\u5f3a\u6548\u5e94<em>\u4e0d\u4f1a\u5ef6\u7eed\u5230\u540e\u7eed\u72ec\u7acb\u4efb\u52a1\u4e2d<\/em>\u3002<\/p>\n<p><\/p>\n<h2>6. \u6700\u4f18\u4eba\u673a\u534f\u4f5c\u7b56\u7565\u7814\u7a76<\/h2>\n<h3>6.1 \u82cf\u683c\u62c9\u5e95\u5f0fAI\u5bfc\u5e08<\/h3>\n<p>\u591a\u9879\u6700\u65b0\u7814\u7a76\u4e00\u81f4\u8868\u660e\uff0c\u5c06LLM\u914d\u7f6e\u4e3a\u82cf\u683c\u62c9\u5e95\u5f0f\u5bfc\u5e08\uff08\u901a\u8fc7\u63d0\u95ee\u5f15\u5bfc\u800c\u975e\u76f4\u63a5\u7ed9\u7b54\u6848\uff09\u53ef\u4ee5\u540c\u65f6\u5b9e\u73b0\u6548\u7387\u63d0\u5347\u548c\u6280\u80fd\u4fdd\u6301\uff1a<\/p>\n<ul>\n<li><strong>SocraticAI (Sunil & Thakkar, 2024)\uff1a<\/strong>\u5c06LLM\u91cd\u6784\u4e3a\u6709\u7ea6\u675f\u7684CS\u6559\u5b66\u5bfc\u5e08\uff0c\u901a\u8fc7\u7ed3\u6784\u5316\u4e92\u52a8\u4fc3\u8fdb\u5b66\u751f\u8868\u8fbe\u63a8\u7406\u8fc7\u7a0b\u3002<\/li>\n<li><strong>STAP (2025)\uff1a<\/strong>\u82cf\u683c\u62c9\u5e95\u5f0f\u81ea\u9002\u5e94\u7f16\u7a0b\u5bfc\u5e08\uff0c\u5c06LLM\u89d2\u8272\u4ece\"\u795e\u8c15\"\u53d8\u4e3a\"\u82cf\u683c\u62c9\u5e95\u5bfc\u5e08\"\u3002<\/li>\n<li><strong>EULER (2024)\uff1a<\/strong>\u5fae\u8c03LLM\u5b9e\u73b0\u82cf\u683c\u62c9\u5e95\u5f0f\u4e92\u52a8\uff0c\u7528\u63d0\u95ee\u5f15\u5bfc\u5b66\u751f\u81ea\u5df1\u53d1\u73b0\u7b54\u6848\u3002<\/li>\n<li><strong>Khan Academy + Khanmigo\uff1a<\/strong>\u5b9a\u5236\u63d0\u793a\u4f7fAI\u4f7f\u7528\u82cf\u683c\u62c9\u5e95\u6cd5\uff0c\u51e0\u4e4e\u6bcf\u8f6e\u5bf9\u8bdd\u90fd\u63d0\u95ee\u3002<\/li>\n<li><strong>RL\u5bf9\u9f50\u6559\u5b66\u6cd5 (2025)\uff1a<\/strong>\u7528\u5f3a\u5316\u5b66\u4e60\u8bad\u7ec3LLM\u5bfc\u5e08\u4f7f\u7528\u82cf\u683c\u62c9\u5e95\u5f0f\u63d0\u95ee\u548c\u9488\u5bf9\u6027\u63d0\u793a\u3002<\/li>\n<\/ul>\n<h3>6.2 \u811a\u624b\u67b6\u5f0fAI\u8f85\u52a9<\/h3>\n<ul>\n<li><strong>DBox (2025, CHI '25)\uff1a<\/strong>\u901a\u8fc7\u5b66\u4e60\u8005-AI\u5171\u540c\u5206\u89e3\u95ee\u9898\u7684\u4ea4\u4e92\u6b65\u9aa4\u6811\uff0c\u652f\u6301\u6784\u601d\u548c\u5b9e\u73b0\u9636\u6bb5\uff0c\u540c\u65f6<em>\u57f9\u517b\u72ec\u7acb\u601d\u7ef4<\/em>\u3002<\/li>\n<li><strong>AI-based scaffolding (2025)\uff1a<\/strong>AI\u811a\u624b\u67b6\u5bf9\u5b66\u4e60\u8005\u7684\u95ee\u9898\u89e3\u51b3\u80fd\u529b\u548c\u5143\u8ba4\u77e5\u610f\u8bc6\u6709\u663e\u8457\u6b63\u5411\u5f71\u54cd\u3002<\/li>\n<li><strong>More AI Assistance Reduces Cognitive Engagement (2025)\uff1a<\/strong>AI\u8f85\u52a9\u6c34\u5e73\u8d8a\u9ad8\uff0c\u8ba4\u77e5\u53c2\u4e0e\u5ea6\u8d8a\u4f4e\u2014\u2014\u5b58\u5728\"AI\u8f85\u52a9\u56f0\u5883\"\u3002<\/li>\n<\/ul>\n<h3>6.3 \u7ed3\u5bf9\u7f16\u7a0b\u4e2d\u7684AI<\/h3>\n<p>Fast and Forgettable (2025)\uff1a\u5bf9\u6bd4AI\u8f85\u52a9\u7ed3\u5bf9\u7f16\u7a0b\u4e0e\u4f20\u7edf\u7ed3\u5bf9\u7f16\u7a0b\u7684\u65b0\u624b\u5b66\u4e60\u6548\u679c\uff0c\u53d1\u73b0AI\u8f85\u52a9\u53ef\u80fd\u66f4\u597d\u4fdd\u5b88\u5730\u4f7f\u7528\uff0c\u4e0e\u4f20\u7edf\u6a21\u5f0f\u7ed3\u5408\u3002<\/p>\n<p><\/p>\n<h2>7. \u7efc\u5408\u5206\u6790\u4e0e\u5b9e\u8df5\u5efa\u8bae<\/h2>\n<div class=\"box\">\n<strong>\u6838\u5fc3\u77db\u76fe\uff1a<\/strong>AI\u63d0\u5347\u77ed\u671f\u6548\u7387\uff0c\u4f46\u4ee5\u957f\u671f\u6280\u80fd\u9000\u5316\u4e3a\u4ee3\u4ef7\u3002\u8fd9\u4e00\u53d1\u73b0\u8d2f\u7a7f\u6240\u6709\u6587\u732e\u2014\u2014\u4ece\u5b9e\u9a8c\u5ba4RCT\u5230\u5927\u89c4\u6a21\u73b0\u573a\u5b9e\u9a8c\u3002\n<\/div>\n<h3>7.1 \u8bc1\u636e\u6c47\u603b\u8868<\/h3>\n<table>\n<tr>\n<th>\u7ef4\u5ea6<\/th>\n<th>AI\u8f85\u52a9\u66f4\u4f18<\/th>\n<th>\u4eba\u7c7b\u72ec\u7acb\u66f4\u4f18<\/th>\n<th>\u8bc1\u636e\u7b49\u7ea7<\/th>\n<\/tr>\n<tr>\n<td>\u5e38\u89c4\/\u5df2\u77e5\u590d\u6742\u95ee\u9898<\/td>\n<td>\u2705 \u6548\u7387+26~55%<\/td>\n<td>\u2014<\/td>\n<td>\u5f3a\uff083\u9879\u5927\u89c4\u6a21RCT\uff09<\/td>\n<\/tr>\n<tr>\n<td>\u524d\u6cbf\u5916\u5f00\u521b\u6027\u95ee\u9898<\/td>\n<td>\u2014<\/td>\n<td>\u2705 \u6b63\u786e\u7387+19%<\/td>\n<td>\u5f3a\uff08BCG RCT\uff09<\/td>\n<\/tr>\n<tr>\n<td>\u72ec\u7acb\u89e3\u9898\u80fd\u529b\uff08\u6280\u80fd\u8f6c\u79fb\uff09<\/td>\n<td>\u2014<\/td>\n<td>\u2705 \u6b63\u786e\u7387+16%<\/td>\n<td>\u5f3a\uff083\u9879RCT, N=1,222\uff09<\/td>\n<\/tr>\n<tr>\n<td>\u5b66\u4e60\u65b0\u6280\u80fd\u7684\u6548\u679c<\/td>\n<td>\u2014<\/td>\n<td>\u2705 \u6982\u5ff5\u7406\u89e3\u66f4\u4f18<\/td>\n<td>\u4e2d\uff08Shen & Tamkin RCT\uff09<\/td>\n<\/tr>\n<tr>\n<td>\u521b\u9020\u6027\u4efb\u52a1\uff08\u4e2a\u4f53\u5c42\u9762\uff09<\/td>\n<td>\u2705 \u521b\u610f\u8bc4\u5206+30~40%<\/td>\n<td>\u2014<\/td>\n<td>\u4e2d<\/td>\n<\/tr>\n<tr>\n<td>\u521b\u9020\u6027\u4efb\u52a1\uff08\u96c6\u4f53\u591a\u6837\u6027\uff09<\/td>\n<td>\u2014<\/td>\n<td>\u2705 \u5185\u5bb9\u591a\u6837\u6027\u66f4\u9ad8<\/td>\n<td>\u4e2d<\/td>\n<\/tr>\n<\/table>\n<h3>7.2 \u5b9e\u8df5\u7b56\u7565\u2014\u2014\"\u534a\u4eba\u9a6c\"\u6a21\u5f0f<\/h3>\n<ol>\n<li><strong>AI\u505a\u57fa\u7840\u7814\u7a76\/\u6a21\u5f0f\u8bc6\u522b\/\u6837\u677f\u4ee3\u7801\u751f\u6210\uff1a<\/strong>\u5229\u7528AI\u7684\u901f\u5ea6\u548c\u5e7f\u5ea6\u4f18\u52bf\u3002<\/li>\n<li><strong>\u4eba\u505a\u67b6\u6784\u5224\u65ad\/\u521b\u65b0\/\u4e0a\u4e0b\u6587\u7406\u89e3\/\u5f02\u5e38\u5904\u7406\uff1a<\/strong>\u4fdd\u7559\u4eba\u7c7b\u72ec\u6709\u7684\u5224\u65ad\u529b\u3002<\/li>\n<li><strong>\u82cf\u683c\u62c9\u5e95\u5f0fPrompt\uff1a<\/strong>\u660e\u786e\u6307\u793aAI\"\u901a\u8fc7\u63d0\u95ee\u5f15\u5bfc\uff0c\u4e0d\u8981\u76f4\u63a5\u7ed9\u7b54\u6848\"\u3002<\/li>\n<li><strong>\u81ea\u6211\u8bc4\u4f30\u673a\u5236\uff1a<\/strong>\u5b9a\u671f\u5728\u65e0AI\u73af\u5883\u4e0b\u6d4b\u8bd5\u72ec\u7acb\u80fd\u529b\uff0c\u9632\u6b62\u6280\u80fd\u9000\u5316\u3002<\/li>\n<li><strong>\u5bf9\u65b0\u624b\u7684\u5173\u952e\u5efa\u8bae\uff1a<\/strong>\u5728\u65f6\u95f4\u538b\u529b\u4e0b\uff0c\u7528AI\u83b7\u53d6\u77e5\u8bc6\u6846\u67b6\u548c\u6587\u732e\u7d22\u5f15\uff0c\u4f46\u6838\u5fc3\u5b9e\u73b0\u548c\u67b6\u6784\u51b3\u7b56\u7531\u81ea\u5df1\u5b8c\u6210\u3002<\/li>\n<\/ol>\n<p><\/p>\n<div class=\"ref-section\">\n<h2>\u53c2\u8003\u6587\u732e<\/h2>\n<p class=\"ref\">Boussioux, L., Lane, J. N., Zhang, M., Vladimirsky, V., & Lakhani, K. R. (2024). The crowdless future? Generative AI and creative problem-solving. <em>Organization Science<\/em>, <em>35<\/em>(5), 1589\u20131607. https:\/\/doi.org\/10.1287\/orsc.2023.18430<\/p>\n<p class=\"ref\">Brynjolfsson, E., Li, D., & Raymond, L. R. (2025). Generative AI at work. <em>The Quarterly Journal of Economics<\/em>, <em>140<\/em>(2), 889\u2013942. https:\/\/doi.org\/10.1093\/qje\/qjae031<\/p>\n<p class=\"ref\">Cui, Z., Demirer, M., Jaffe, S., Musolff, L., Peng, S., & Salz, T. (2025). The effects of generative AI on high-skilled work: Evidence from three field experiments. <em>Management Science<\/em>. Advance online publication. https:\/\/doi.org\/10.1287\/mnsc.2025.00535<\/p>\n<p class=\"ref\">Dell'Acqua, F., McFowland, E., III, Mollick, E. R., Lifshitz-Assaf, H., Kellogg, K., Rajendran, S., Krayer, L., Candelon, F., & Lakhani, K. R. (2025). Navigating the jagged technological frontier: Field experimental evidence of the effects of AI on knowledge worker productivity and quality. <em>Organization Science<\/em>. Advance online publication. https:\/\/doi.org\/10.1287\/orsc.2025.21838<\/p>\n<p class=\"ref\">Doshi, A. R., & Hauser, O. (2024). Generative AI enhances individual creativity but reduces the collective diversity of novel content. <em>Science Advances<\/em>, <em>10<\/em>(29), eadn5290. https:\/\/doi.org\/10.1126\/sciadv.adn5290<\/p>\n<p class=\"ref\">Hohenstein, J., Kizilcec, R. F., DiGiacomo, D., Aghajari, Z., Mubarrat, S., Biehl, M., & Jung, M. F. (2024). Does using artificial intelligence assistance accelerate skill decay and hinder skill development without performers' awareness? <em>Cognitive Research: Principles and Implications<\/em>, <em>9<\/em>, Article 46. https:\/\/doi.org\/10.1186\/s41235-024-00572-8<\/p>\n<p class=\"ref\">Lee, H. P., Sarkar, A., Tankelevitch, L., Drosos, I., Rintel, S., Banks, R., & Wilson, N. (2025). The impact of generative AI on critical thinking: Self-reported reductions in cognitive effort and confidence effects from a survey of knowledge workers. In <em>Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems<\/em> (CHI '25). ACM. https:\/\/doi.org\/10.1145\/3706598.3713778<\/p>\n<p class=\"ref\">Liu, G., Christian, B., Dumbalska, T., Bakker, M. A., & Dubey, R. (2025). AI assistance reduces persistence and hurts independent performance. <em>arXiv preprint arXiv:2604.04721<\/em>. https:\/\/arxiv.org\/abs\/2604.04721<\/p>\n<p class=\"ref\">Noy, S., & Zhang, W. (2023). Experimental evidence on the productivity effects of generative artificial intelligence. <em>Science<\/em>, <em>381<\/em>(6654), 187\u2013192. https:\/\/doi.org\/10.1126\/science.adh2586<\/p>\n<p class=\"ref\">Peng, S., Kalliamvakou, E., Cihon, P., & Demirer, M. (2023). The impact of AI on developer productivity: Evidence from GitHub Copilot. <em>arXiv preprint arXiv:2302.06590<\/em>. https:\/\/arxiv.org\/abs\/2302.06590<\/p>\n<p class=\"ref\">Shen, J. H., & Tamkin, A. (2026). How AI impacts skill formation: Evidence from software development learning. <em>arXiv preprint arXiv:2601.20245<\/em>. https:\/\/arxiv.org\/abs\/2601.20245<\/p>\n<p class=\"ref\">Vaccaro, M., Almaatouq, A., & Malone, T. W. (2024). When combinations of humans and AI are useful: A systematic review and meta-analysis. <em>Nature Human Behaviour<\/em>, <em>8<\/em>, 2295\u20132308. https:\/\/doi.org\/10.1038\/s41562-024-02024-1<\/p>\n<h3 style=\"margin-top: 24px;\">\u8865\u5145\u6587\u732e\u2014\u2014\u6700\u4f18\u534f\u4f5c\u7b56\u7565<\/h3>\n<p class=\"ref\">Chowdhury, A., Roy, S., & Radev, D. (2024). MWPTutor: LLM-powered tutoring with hierarchical guardrails. In <em>Proceedings of the AAAI Conference on Artificial Intelligence<\/em>.<\/p>\n<p class=\"ref\">Dhillon, H., Gu, Z., & Shneiderman, B. (2024). More AI assistance reduces cognitive engagement: Examining the AI assistance dilemma in AI-supported note-taking. <em>arXiv preprint arXiv:2509.03392<\/em>. https:\/\/arxiv.org\/abs\/2509.03392<\/p>\n<p class=\"ref\">Gatti, G., & colleagues. (2024). EULER: Fine tuning a large language model for Socratic interactions. In <em>Proceedings of AIxEDU 2024<\/em> (CEUR Workshop Proceedings, Vol. 3879).<\/p>\n<p class=\"ref\">Kazemitabaar, M., Chow, J., Ma, C. K. T., Ericson, B. J., Weintrop, D., & Grossman, T. (2025). Fast and forgettable: A controlled study of novices' performance, learning, workload, and emotion in AI-assisted and human pair programming paradigms. <em>arXiv preprint arXiv:2604.18538<\/em>. https:\/\/arxiv.org\/abs\/2604.18538<\/p>\n<p class=\"ref\">Liu, J., Huang, Z., & colleagues. (2024). SocraticLM: Exploring Socratic personalized teaching with large language models. In <em>Advances in Neural Information Processing Systems<\/em> (NeurIPS 2024).<\/p>\n<p class=\"ref\">Sunil, S., & Thakkar, P. (2024). SocraticAI: Transforming LLMs into guided CS tutors through scaffolded interaction. <em>arXiv preprint arXiv:2512.03501<\/em>. https:\/\/arxiv.org\/abs\/2512.03501<\/p>\n<p class=\"ref\">Wang, Z., Liu, Y., & Chen, X. (2025). STAP: A Socratic tutor for adaptive programming with pedagogical scaffolding. In <em>Proceedings of the 56th ACM Technical Symposium on Computer Science Education<\/em> (SIGCSE TS 2025). ACM. https:\/\/doi.org\/10.1145\/3775073.3775165<\/p>\n<p class=\"ref\">Yang, K., Zhang, T., & Li, S. (2025). From problem-solving to teaching problem-solving: Aligning LLMs with pedagogy using reinforcement learning. <em>arXiv preprint arXiv:2505.15607<\/em>. https:\/\/arxiv.org\/abs\/2505.15607<\/p>\n<p class=\"ref\">Zhang, L., Ma, N., & Zhao, G. (2024). Generative AI in education: From foundational insights to the Socratic playground for learning. <em>arXiv preprint arXiv:2501.06682<\/em>. https:\/\/arxiv.org\/abs\/2501.06682<\/p>\n<p class=\"ref\">Zhao, Y., Wu, R., & Peng, H. (2025). DBox: Scaffolding algorithmic programming learning through learner-LLM co-decomposition. In <em>Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems<\/em> (CHI '25). ACM. https:\/\/doi.org\/10.1145\/3706598.3713748<\/p>\n<p class=\"ref\">(2025). Human-generative AI collaboration enhances task performance but does not persist in subsequent tasks. <em>Scientific Reports<\/em>. https:\/\/doi.org\/10.1038\/s41598-025-98385-2<\/p>\n<p class=\"ref\">(2025). AI-assisted vs human-only evidence review: Results from a comparative study. UK Government Behavioural Insights Team. https:\/\/www.gov.uk\/government\/publications\/ai-assisted-vs-human-only-evidence-review<\/p>\n<p class=\"ref\">(2026). Generative AI adoption in human creative tasks: Experimental evidence. <em>Journal of Economic Behavior & Organization<\/em>. https:\/\/doi.org\/10.1016\/j.jebo.2026.01.002<\/p>\n<p class=\"ref\">(2025). Can pedagogical agent-based scaffolding boost information problem solving? <em>Education and Information Technologies<\/em>. https:\/\/doi.org\/10.1007\/s10639-025-13784-2<\/p>\n<p class=\"ref\">(2025). The effect of AI-based scaffolding on problem solving and metacognitive awareness in learners. <em>ResearchGate<\/em>. https:\/\/doi.org\/10.13140\/RG.2.2.33921.47201<\/p>\n<\/div>\n<p style=\"text-align:center;margin-top:40px;color:#999;font-size:0.85em;\">\n\u8986\u76d6\u6587\u732e\u6570\uff1a27\u7bc7 \u00b7 \u7c7b\u578b\uff1a\u5143\u5206\u6790\/\u7cfb\u7edf\u7efc\u8ff0\/RCT\/\u73b0\u573a\u5b9e\u9a8c\/\u7406\u8bba\u5206\u6790 \u00b7 \u8de8\u5ea6\uff1a2023\u20132026\n<\/p>\n<\/div>\n<p><\/body><br \/>\n<\/html><\/p>\n","protected":false},"excerpt":{"rendered":"<p>AI\u8f85\u52a9 vs \u4eba\u7c7b\u72ec\u7acb\u89e3\u51b3\u95ee\u9898\u6548\u7387\u5bf9\u6bd4\u2014\u2014\u6587\u732e\u7efc\u8ff0 AI\u8f85\u52a9\u4e0e\u4eba\u7c7b\u72ec\u7acb\u89e3\u51b3\u95ee\u9898\u6548\u7387\u5bf9\u6bd4 \u2014\u2014\u805a\u7126\u590d\u6742\/\u5f00\u521b\u6027\u95ee\u9898\u7684\u6587\u732e\u7efc\u8ff0 \u6838\u5fc3\u95ee\u9898\uff1a\u5bf9\u4e8e\u56f0\u96be\u7684\u5f00\u521b\u6027\u95ee\u9898\uff08\u5982\u5b9a\u5236\u4e0b\u8f7d\u534f\u8bae\u3001\u5e26\u5bbd\u901f\u7387\u9650\u5236\u7b49\uff09\uff0c\u4eba\u7c7b\u5229\u7528AI\u8f85\u52a9\u4e0e\u901a\u8fc7","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"emotion":"","emotion_color":"","title_style":"","license":"","footnotes":""},"categories":[1],"tags":[],"class_list":["post-116","post","type-post","status-publish","format-standard","hentry","category-uncategorized"],"_links":{"self":[{"href":"https:\/\/www.tom-thu.cn\/index.php?rest_route=\/wp\/v2\/posts\/116","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.tom-thu.cn\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.tom-thu.cn\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.tom-thu.cn\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.tom-thu.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=116"}],"version-history":[{"count":1,"href":"https:\/\/www.tom-thu.cn\/index.php?rest_route=\/wp\/v2\/posts\/116\/revisions"}],"predecessor-version":[{"id":117,"href":"https:\/\/www.tom-thu.cn\/index.php?rest_route=\/wp\/v2\/posts\/116\/revisions\/117"}],"wp:attachment":[{"href":"https:\/\/www.tom-thu.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=116"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.tom-thu.cn\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=116"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.tom-thu.cn\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=116"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}