[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-135137-en":3,"doc-seo-135137-105":30,"detail-sidebar-cat-0-en-105":91},{"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":20,"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":27,"seo_description":14,"update_tm":28,"read_time":29},135137,549758146520,"Patrick","https://ap-avatar.wpscdn.com/avatar/80002397d8c0411e94?_k=1775819394049821470",8,"Research & Report","Appendix B - Bot Detection Method Details","Appendix B details multiple bot-detection approaches used to identify automated or fraudulent responses in an online survey setting. It covers reCAPTCHA v2/v3 behavior-based scoring, reverse shibboleth questions targeting human inability, direct honesty checks, grid-based attention tasks, and logo-based image recognition. The appendix further describes consistency checks, honeypot and prompt-injection traps, paradata collection for device and IP patterns, and copy-paste detection for open-ended answers, supported by prior research references.","Appendix B: Bot Detection Method Details\nreCAPTCHA v2 and v3: reCAPTCHA v2 asks users to click a box that says “I am not a robot” and v3 provides a bot score without user interaction. Both tools purport to identify bots using information about mouse movements, browser metadata, and cookies. reCAPTCHA is used extensively across the internet to detect bots (e.g., see Von Ahn, Blum, and Langford 2004).\n“Reverse Shibboleth” questions: Shibboleth questions are questions that a target audience (e.g., Physicians) can easily complete, but respondents outside the target audience cannot. Reverse shibboleth questions are the reverse—questions that the target audience (in this case humans) cannot easily answer but those outside the target audience can. For example, if we ask respondents to “please write a ‘Hello World’ program in Fortran,” most human respondents will be unable to comply, but LLMs can easily answer the question. Our survey included three reverse shibboleth questions.\nDirect honesty test: We directly asked Operator if it was a bot or a real, live human.\nGrid-based attention check: We asked respondents how often they did certain activities in the past 12 months. One of the activities was “flew to the moon.” Respondents who are paying attention should select ‘0 times.’\nImage recognition: We showed respondents two questions with logos associated a brand and asked the respondent to report which brand is associated with the displayed logo. The first question displayed the Fanta logo and included four response options. The second displayed the Target logo and asked the respondent to write the name of the associated brand in a text box.\nConsistency checks: We asked three pairs of similar questions, one near the beginning of the survey and once near the end of the survey. The first pair asked respondents to report their age and then their year of birth. The second pair asked respondents to report their income and then to select one of several income categories. The third asked respondents to report whether they have any children and then to report the number of children they have. These types of consistency checks are widely recommended as a way to detect inattentive or fraudulent responses (Meade and Craig 2012; Zhang et al. 2022)\nHoneypot question: We included a question asking respondents to select the highest number. The correct answer was hidden using javascript. In theory, if a bot were parsing the HTML of the survey page, it could see and select the hidden correct answer. This method has been used by others with mixed results (see, e.g., Storozuk et al. 2020).\nPrompt injection question: We included a question asking respondents to report their favorite ice cream flavor. In very small white text, we asked respondents to use the term “flavor profile” in their response. In theory, human respondents won’t see the additional instruction, but a bot parsing the page’s HTML or someone copying the question text into an LLM would. This method is based on a technique Joshua McCrain used to catch students cheating on college essays as described in a Wall Street Journal report.\nParadata collection: We collected a number of different types of respondent paradata including device characteristics like browser type and version, device time zone, IP address, and more. Certain paradata patterns have been shown to be a strong predictor of fraudulent activity (Teitcher et al. 2015).\nCopy-paste detection: For each open-ended question, we included JavaScript code to record whether the respondent had pasted text onto the page. Veselovsky, Ribeiro, and West (2023) suggest copy-paste detection is a reliable way to detect LLM usage in text production tasks.","cbCaigrE3tjvPoUh","https://ap.wps.com/l/cbCaigrE3tjvPoUh","docx",30118,1,2,"English","en",105,"# Method Overview\n## CAPTCHA-Based Detection\n## Question-Based Traps and Checks\n## Paradata and Response Forensics","[{\"question\":\"How do reCAPTCHA v2 and v3 help detect bots?\",\"answer\":\"reCAPTCHA v2 uses a user-interaction task (“I am not a robot”), while v3 returns a bot score without user interaction. Both infer bot likelihood using signals such as mouse behavior, browser metadata, and cookies.\"},{\"question\":\"What are reverse shibboleth questions and why include them?\",\"answer\":\"Reverse shibboleth questions are designed so humans cannot answer easily, while bots may generate responses. The appendix notes that the survey included three such questions to distinguish respondents based on capability.\"},{\"question\":\"What role do consistency checks and paradata collection play?\",\"answer\":\"Consistency checks compare paired answers (e.g., age vs. birth year, income vs. categories) to reveal inattentive or fraudulent responding. Paradata collection tracks device and network signals like browser characteristics, time zone, and IP patterns that can predict fraudulent activity.\"}]","Appendix B - Bot Detection Method Details | DOCX",1787305556,5,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"appendix-b-bot-detection-method-details","",{"@graph":36,"@context":85},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,47,50],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":21},"https://docshare.wps.com/document/","Document",{"item":48,"name":12,"@type":43,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":43,"position":52},"https://docshare.wps.com/document/appendix-b-bot-detection-method-details/135137/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/vnd.openxmlformats-officedocument.wordprocessingml.document","2026-08-23","2026-08-21",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"How do reCAPTCHA v2 and v3 help detect bots?","Question",{"text":75,"@type":76},"reCAPTCHA v2 uses a user-interaction task (“I am not a robot”), while v3 returns a bot score without user interaction. Both infer bot likelihood using signals such as mouse behavior, browser metadata, and cookies.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What are reverse shibboleth questions and why include them?",{"text":80,"@type":76},"Reverse shibboleth questions are designed so humans cannot answer easily, while bots may generate responses. The appendix notes that the survey included three such questions to distinguish respondents based on capability.",{"name":82,"@type":73,"acceptedAnswer":83},"What role do consistency checks and paradata collection play?",{"text":84,"@type":76},"Consistency checks compare paired answers (e.g., age vs. birth year, income vs. categories) to reveal inattentive or fraudulent responding. Paradata collection tracks device and network signals like browser characteristics, time zone, and IP patterns that can predict fraudulent activity.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,109,114,119,122,127,130,134],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":21,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":29,"doc_module":4,"doc_module_name":46,"category_name":106,"show_sort_weight":107,"slug":108},"Comic",60,"comic",{"id":110,"doc_module":4,"doc_module_name":46,"category_name":111,"show_sort_weight":112,"slug":113},6,"Technology",50,"technology",{"id":115,"doc_module":4,"doc_module_name":46,"category_name":116,"show_sort_weight":117,"slug":118},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":46,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":46,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":46,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":46,"category_name":136,"show_sort_weight":29,"slug":137},19,"General","general"]