IIT Delhi's ETD-2026 Curtain Raiser | ETDs in the Age of AI: skim's analysis identifies 32 key moments, with 1 potential conflict of interest flagged. IIT Delhi's Curtain Raiser for ETD-2026 convenes academic leaders, AI researchers, and library directors to examine how artificial intelligence transforms electronic theses, scholarly discovery, and research integrity.
Category: Education. Format: Panel Discussion. YouTube video analyzed by skim.
Key Points (32)
1. Ratna Das Opens the ETD-2026 Curtain Raiser
Timestamp: 00:05:58 to 00:09:28
Ratna Das welcomes attendees to IIT Delhi for the Curtain Raiser of the ETD-2026 international symposium, Annual Library Day, and the third CLAN alumni meet. She notes that ETD-2026 will run from October 23 to 25, 2026 in collaboration with NDLTD, and announces that abstract submissions remain open until May 30.
Significance (Low): Sets the institutional scope and timeline for India's third international electronic theses symposium, aligning open knowledge access with artificial intelligence.
Sources in support: Ratna Das (Senior Library Information Assistant, Central Library, IIT Delhi)
2. A professor Welcomes National Delegates
Timestamp: 00:18:13 to 00:27:23
Ratna Das introduces Professor and felicitate visiting dignitaries with live plants before Professor Alappat delivers the welcome address. Speaking on behalf of the Central Library family, he welcomes delegates to the 29th international symposium curtain raiser and previews the new library website launch.
Significance (Low): Demonstrates cross-institutional backing from IGNOU, IGNCA, DRDO, DELNET, and Clarivate for IIT Delhi's library modernization initiatives.
Sources in support: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers), Ratna Das (Senior Library Information Assistant, Central Library, IIT Delhi)
3. A chair Traces India's ETD Movement
Timestamp: 00:27:30 to 00:35:23
A chair outlines the history of the ETD movement from Edward Fox establishing NDLTD in 1996 to the creation of India's Shodhganga repository following a 2005 UGC meeting. He reports that Shodhganga holds over five lakh open-access theses, making India the largest open ETD contributor after the United States, while global metadata records exceed six million.
Significance (High): Establishes how regulatory mandates and INFLIBNET infrastructure propelled India to become the world's second-largest producer of open-access doctoral theses.
Sources in support: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers)
4. Integrating Indian Knowledge Systems and AI Ethics
Timestamp: 00:35:23 to 00:43:45
Continuing his address, a chair highlights India's One Nation One Subscription initiative and the 2018 UGC plagiarism regulations before calling for ETD-2026 to address ethical AI and Indian Knowledge Systems. He cites Gyan Bharatam identifying 75 lakh manuscripts, IGNCA digitizing over six lakh manuscripts, and a three-state survey showing 95 percent of residents in Arunachal, Sikkim, and Manipur rely on traditional healing.
Significance (Medium): Broadens the symposium's mandate beyond library cataloging to encompass AI research ethics, indigenous manuscript digitization, and Asia-Pacific leadership within NDLTD.
Sources in support: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers)
5. Kiran Chauhan on Defense Library Relevance
Timestamp: 00:43:50 to 00:46:05
Kiran Chauhan, Director of DESIDOC at DRDO, addresses the gathering on behalf of defense sector libraries serving scientists and technologists across DRDO laboratories. She affirms that examining electronic theses, emerging tools, and research ethics in the age of AI is vital for specialized scientific libraries.
Significance (Low): Confirms that national defense research documentation centers face the same AI tooling and research ethics challenges as university libraries.
Sources in support: Kiran Chauhan (Director, DESIDOC, DRDO)
6. A speaker Unveils ETD-2026 Portal and Ideathon
Timestamp: 00:46:05 to 00:54:20
After the guest digitally inaugurates the ETD-2026 portal, a speaker details the extended May 30 abstract deadline, submission tracks, and a new two-to-three-hour Ideathon where teams build prototypes for library problems using any AI tool. He also announces early-bird registration at 5,900 rupees until June 30, rising to 7,080 rupees thereafter, alongside sponsors Clarivate and VSoft Tech.
Significance (Medium): Introduces hands-on rapid prototyping with generative AI tools into an international library science conference while setting concrete participation deadlines.
Sources in support: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers), Ratna Das (Senior Library Information Assistant, Central Library, IIT Delhi)
7. A speaker Launches Django-Based Library Website
Timestamp: 00:54:23 to 00:58:48
A speaker presents and inaugurates IIT Delhi Central Library's new website, built with team members Sudhanshu and Nikhil Raj under Head Librarian Nabi Hasan. He explains that the platform uses the Python Django framework with role-based access control and APIs that unify OpenAlex, the library management system, DSpace, and chatbot features.
Significance (Medium): Demonstrates how academic libraries can replace fragmented portals with an open-source Python architecture that unifies repository and bibliometric APIs.
Sources in support: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers)
8. A chair Presents Best Library User Awards
Timestamp: 01:04:10 to 01:10:05
Following an AI-generated video celebrating the CLAN intern network, a chair announces the Best Library User Awards to recognize data-driven resource utilization across IIT Delhi degree programs. A director presents 5,000-rupee prizes, certificates, and mementos to PhD winner Miriyam, Master's winner Chitesh Kumar Bahuguna via representative, and undergraduate winner Tamojit Das.
Significance (Low): Uses quantitative usage metrics and financial incentives to encourage active student engagement with costly institutional library subscriptions.
Sources in support: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers), Ratna Das (Senior Library Information Assistant, Central Library, IIT Delhi)
9. Over 1,000 Campus Photos Added to Wikimedia Commons
Timestamp: 01:10:09 to 01:17:17
A speaker details the GLAM IIT Delhi collaboration with the Wikimedia community and Wikimedians in residence Aisha and Faiz to document campus heritage under Indian freedom of panorama provisions. During the February 2 to 17, 2026 contest, 75 unique users uploaded over 1,000 photographs to Wikimedia Commons, with top contributor Arya Upamanyu uploading 92 photos.
Significance (Medium): Illustrates a replicable model for universities to crowdsource open-licensed visual archives on Wikimedia Commons while respecting copyright law.
Sources in support: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers), Ratna Das (Senior Library Information Assistant, Central Library, IIT Delhi)
10. A director on Filtering AI Search Overload
Timestamp: 01:17:17 to 01:26:15
Delivering the presidential address, a director contrasts his PhD days at IIT Bombay rushing to physical library desks at 7:30 AM with today's instant AI and data-mining tools. He warns that when a scholar wants 10 papers, AI tools return 10,000, making natural intelligence and skilled librarians essential to filter what is truly relevant and innovative.
Significance (Medium): Frames the core modern research bottleneck not as information scarcity, but as severe retrieval overload requiring human judgment and library curation.
Sources in support: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers)
11. The guest Redefines Libraries as Innovation Hubs
Timestamp: 01:26:18 to 01:34:38
IGNOU Vice Chancellor the guest praises IIT Delhi's library internship program aligned with NEP 2020 and argues that libraries must evolve from silent book racks into living spaces, learning commons, and makerspaces. She urges higher education institutions to treat libraries not as passive service providers, but as strategic academic partners and innovation hubs.
Significance (Medium): Challenges university leadership to reallocate physical and digital library spaces toward collaborative makerspaces and mandatory professional internships.
Sources in support: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers)
12. Librarians as Knowledge Architects Using Agentic AI
Timestamp: 01:34:38 to 01:42:48
Continuing her address, the guest states that digitized theses become living nodes in a global network and proposes deploying agentic AI platforms to recommend research pathways, detect knowledge gaps, and connect interdisciplinary insights automatically. She urges librarians to become knowledge architects supporting last-mile digital access for Viksit Bharat 2047.
Significance (High): Articulates a concrete technical vision for agentic AI in repositories: moving beyond keyword search to automated gap detection and interdisciplinary synthesis.
Sources in support: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers)
13. Mausam Warns Against Indiscriminate AI in Education
Timestamp: 01:42:48 to 01:50:06
Introduced as Founding Head of IIT Delhi's Yardi School of AI, Professor Mausam rejects three extreme reactions to LLMs—denial, doom, and lazy overreliance—in favor of a realist approach. He warns against giving students indiscriminate access to AI because education is about the learning process, whereas AI optimizes assessment outcomes at the expense of student learning. He also notes new Nature papers debating autonomous AI scientists.
Significance (High): Provides a sharp pedagogical warning from a leading AI scientist: optimizing assignment outcomes via LLMs hollows out the cognitive process of education.
Sources in support: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers)
14. Three Paradigms of LLM Information Retrieval
Timestamp: 01:50:06 to 01:57:48
Mausam contrasts three paradigms for getting information to users: LLM as a know-it-all, Retrieval-Augmented Generation (RAG), and LLM as a semantic parser over knowledge bases. He explains that know-it-all models hallucinate citations, guzzle electricity, and cannot easily add or delete facts, whereas RAG adds provenance and semantic parsers translate queries into fast, lightweight SPARQL or SQL programs.
Significance (High): Gives library and enterprise architects a clear trade-off matrix between raw LLMs, RAG pipelines, and structured semantic parsing over knowledge bases.
Sources in support: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers)
15. Fusing LLMs and Supervised ML for Unanswerable Queries
Timestamp: 01:57:48 to 02:07:44
Mausam explains that AI models fail to admit when they do not know an answer, often using execution-guided filtering to force non-empty answers from incomplete knowledge bases. He presents his team's FUSIC model, which combines LLM syntax knowledge with supervised machine learning for proprietary database schemas, and F-FUSIC, which samples multiple answers to detect diffused confidence distributions and abstain on unanswerable queries.
Significance (High): Demonstrates how sampling answer distributions and pairing traditional supervised ML with LLMs prevents hallucinated database queries and enables truthful abstention.
Sources in support: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers)
16. Decomposing Complex Analytics into SQL and Python
Timestamp: 02:07:44 to 02:12:35
Addressing complex analytical reasoning over structured data, Mausam introduces the StarQA dataset of 362 handwritten questions across movies, sports, and e-commerce. Because LLMs make frequent errors writing monolithic SQL for complex reasoning, his team built a pipeline that decomposes questions into text-to-SQL for data fetching followed by LLM-generated Python code, boosting performance from 35 percent to 44 percent.
Significance (High): Shows that splitting data retrieval (SQL) from procedural reasoning (Python) significantly improves LLM accuracy on complex analytical queries, though visual reasoning remains weak.
Sources in support: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers)
17. Hallucinated Citations and the Risk of Weak Graduates
Timestamp: 02:12:35 to 02:18:35
During the audience Q&A, Mausam confirms that standard LLMs fabricate plausible-looking paper titles and author lists unless grounded in RAG or search links. Responding to an anecdote where ChatGPT told a finance researcher to visit a library, he warns that non-experts cannot spot subtle LLM errors, risking a catastrophic pipeline of weaker students entering industry as entry-level jobs disappear.
Significance (Medium): Highlights the compounding workforce risk when students bypass foundational learning with LLMs just as automation eliminates entry-level training roles.
Sources in support: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers)
18. AI Apprenticeship Model and Source Hygiene
Timestamp: 02:18:35 to 02:27:10
Answering librarian questions on future jobs and reliable AI tools, Mausam cites contrasting forecasts from Bill Gates, Vinod Khosla, Geoffrey Hinton, and the World Economic Forum before predicting an apprenticeship model over the next 10 to 15 years where smaller human teams partner with AI. He urges librarians and users to treat source verification as mandatory hygiene to avoid collective AI information bubbles.
Significance (Medium): Advises professionals that while AI will shrink routine teams over the next decade, verifying primary documents remains the essential safeguard against algorithmic echo chambers.
Sources in support: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers), Nabi Hasan (Head Librarian, Central Library, IIT Delhi)
Neutral sources: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers)
19. A director Examines Global Thesis Production
Timestamp: 02:50:15 to 02:58:41
After Nabi Hasan announces lunch logistics, a director notes that predictions of the death of libraries have recurred since DOS and CD-ROMs in 1996, though AI now questions universities themselves. Focusing on information discovery, he highlights surging part-time PhD enrollments in private Indian universities under NIRF data and asks whether a complete global catalog of theses exists.
Significance (Medium): Identifies a foundational gap in scholarly communication: despite booming doctoral enrollments, global bibliographic control over theses remains fragmented.
Sources in support: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers), Nabi Hasan (Head Librarian, Central Library, IIT Delhi)
20. Declining Thesis Citations in Web of Science
Timestamp: 02:58:41 to 03:05:51
A director examines how often theses are actually cited in scholarly literature, citing a 2008 study by Vincent Larivière showing thesis citations in Web of Science peaked at just 1 percent in 1940-1950 before falling to 0.5 percent. Updating the query through recent years across roughly 900,000 cited thesis references in Web of Science, Madhan shows thesis citations continue to decline, with only about 270 theses cited at least 100 times.
Significance (High): Provides hard scientometric evidence that doctoral theses represent a tiny and shrinking fraction of cited references in mainstream journal literature.
Sources in support: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers)
21. Historic Pre-1950 Theses Versus Commercial Silos
Timestamp: 03:05:51 to 03:13:36
Contrasting quantitative citation decline with qualitative impact, a director notes that Web of Science still records citations to 19th-century theses from 1892, as well as landmark pre-1950 dissertations by Albert Einstein and Claude Shannon. However, when searching Google for influential dissertations, users are directed to ProQuest, a commercial database holding six million theses that charges students roughly 100 dollars for open-access hosting.
Significance (Medium): Highlights how pre-1950 theses shaped foundational science and exposes how search engines privilege commercial aggregators over national open repositories.
Sources in support: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers)
22. Shodhganga's Metadata and Harvesting Gaps in OpenAlex
Timestamp: 03:13:36 to 03:22:48
A director reveals that while OpenAlex indexes over 8 million dissertations and OATD indexes 7.4 million, India's Shodhganga repository sits largely in a silo. Out of more than 600,000 theses in Shodhganga, OpenAlex has scraped only about 100,000 records and OATD only around 35,000, with many scraped records suffering from missing abstracts, false closed-access tags, and date typos like 1882 instead of 1982.
Significance (High): Exposes a critical operational failure in Indian open access: storing 600,000 theses on DSpace achieves little global impact if broken metadata blocks OAI-PMH harvesting.
Sources in support: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers)
23. The panelist Details Delhi University's Digital Archive
Timestamp: 03:22:48 to 03:33:34
Opening the panel moderated by a chair, the panelist reveals that Delhi University digitized 60 lakh pages and 8,000 theses in 2008, but the entire collection vanished because it lacked institutional policy.
Significance (High): Offers a sobering lesson on digital preservation failure—losing 6 million digitized pages—and provides a comprehensive four-phase institutional blueprint for AI-native archiving.
Sources in support: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers)
24. A director Warns AI Misses Non-Digital Scholarship
Timestamp: 03:33:34 to 03:42:25
A director outlines DELNET's AI strategy around the four Cs—content, connectivity, collaborations, and clients—calling for AI knowledge extraction to identify research gaps for academia-industry partnerships and decentralized shared AI infrastructure, noting India's 10,300-crore AI mission. Highlighting DELNET's AI Mission, she warns researchers that because not more than 5 percent of world content is digital, relying solely on AI tools misses 70 to 80 percent of published works.
Significance (High): Counters the assumption that LLM literature reviews are comprehensive by reminding scholars that vast print archives remain invisible to AI crawlers.
Sources in support: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers)
25. A director Defends ProQuest Curation and Privacy
Timestamp: 03:42:25 to 03:49:25
Responding to a chair and a director, Clarivate Senior Director argues that ProQuest remains the gold standard since the 1930s because it provides metadata consistency, subject classification, download dashboards, and indexing in Web of Science across 10,000 institutions. She states that Clarivate never uses thesis data to train LLMs and clarifies that depositing an electronic copy in ProQuest is free, with a one-time fee charged only for optional open-access hosting.
Significance (Medium): Presents the commercial publisher counterargument: institutions pay for curated metadata, citation integration, and contractual guarantees against LLM training.
Sources in support: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers)
Neutral sources: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers)
26. Centralized Silos and NDLTD's Cloud Hosting Crunch
Timestamp: 03:49:25 to 03:55:04
A chair notes that INFLIBNET has repeatedly declined NDLTD's requests to share Shodhganga metadata for Global ETD Search. A director explains that because Shodhganga and OATD both use OAI-PMH, sharing metadata requires zero extra cost, but centralization allows one authority's refusal to block hundreds of Indian universities. Gaur then reveals that NDLTD faces a funding crunch and invites institutions with cloud capacity to host Global ETD Search.
Significance (High): Demonstrates how administrative centralization can bottleneck open-access interoperability and reveals that even major global metadata aggregators face cloud hosting deficits.
Sources in support: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers)
27. The panelist Proposes Mandatory AI Disclosure Logs
Timestamp: 03:55:04 to 04:02:29
Asked how libraries should handle AI-paraphrased plagiarism, the panelist argues that originality is the presence of authentic human contribution rather than the absence of AI. He proposes classifying research into four AI-use tiers, shifting from detection to mandatory disclosure via prompt diaries and AI use logs, strengthening PhD viva voce exams, appointing institutional AI Integrity Officers, and noting that Turnitin cannot reliably separate language editing from generation.
Significance (High): Provides a concrete, actionable governance blueprint for universities to replace unreliable AI text detectors with transparent process logs and oral defense rigor.
Sources in support: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers)
28. Revising India's 2018 Plagiarism Policy for the AI Era
Timestamp: 04:02:29 to 04:11:55
In rapid closing statements, a director proposes an apex regulatory body for research ethics and researcher integrity oaths modeled on Bhutan and medical ethics, a director stresses transparent publisher AI policies, and a director calls for AI-assisted bibliographic control. Concluding the panel, Ramesh C.
Significance (High): Signals an upcoming regulatory push to update India's national 2018 UGC plagiarism rules with mandatory AI disclosure requirements.
Sources in support: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers)
29. Ideathon Warm-Up Tackles Real-Time Seat Occupancy
Timestamp: 04:12:30 to 04:21:00
A speaker and Charu open the Ideathon warm-up quiz with jury members Vijay, a speaker, and Manu to test practical library solutions under a two-minute limit. Addressing real-time library seat occupancy, PhD student Abhilash proposes applying image processing to existing CCTV feeds to display green/red seat maps on entrance dashboards, while another participant suggests IoT thermal seat sensors.
Significance (Low): Contrasts low-cost computer vision over existing CCTV cameras with hardware-intensive IoT thermal sensors for solving campus library crowding.
Sources in support: Abhilash (PhD Student in Computational Biology, IIT Delhi), Charu (Ideathon Co-Moderator, Central Library, IIT Delhi), YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers)
Neutral sources: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers)
30. RFID Handhelds and Machine Learning for Shelf Accuracy
Timestamp: 04:21:01 to 04:30:28
For book navigation and misshelved book detection, a speaker describes using RFID handheld readers with accession numbers at Central Sanskrit University, while Ajay of IIM Kashipur critiques expensive thermal sensors and advocates DDC OPAC rack mapping and section zoning. Arvind of Sai University proposes using RFID shelf readers on wooden racks or training machine learning models on daily smartphone photos of book spine labels to flag misplaced call numbers.
Significance (Low): Highlights practical trade-offs in stack management between metal-rack RFID interference and mobile computer-vision spine scanning.
Sources in support: Arvind (Library Manager, Sai University, Chennai), YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers)
Neutral sources: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers)
31. AI Metadata Extraction and Ideathon Winner Awards
Timestamp: 04:30:28 to 04:41:45
Addressing LMS migration errors and AI in the ETD lifecycle, Srijan from the University of Calcutta proposes LLM feature extraction during data import, while a speaker suggests Google Books API reconciliation.
Significance (Low): Synthesizes practical library AI use cases: automated metadata reconciliation during software migration, BERT metadata extraction, and multilingual abstract translation.
Sources in support: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers)
Neutral sources: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers), Charu (Ideathon Co-Moderator, Central Library, IIT Delhi)
32. A speaker Delivers Concluding Vote of Thanks
Timestamp: 04:41:46 to 04:48:40
Ratna Das invites attendees to scan a five-question QR feedback poll to shape the October symposium before Deputy Librarian a speaker summarizes the morning's milestones. Chavan thanks Chief Guest, a director, a professor, a chair, keynote speakers, panelists, CLAN alumni, and staff, reminding delegates of the May 30 paper submission deadline and the October 23-25 ETD-2026 conference.
Significance (Low): Closes the inaugural curtain raiser by consolidating institutional partnerships and directing participants toward the May 30 submission deadline.
Sources in support: YouTube: IIT Delhi (Host Channel and Unidentified Audience Speakers), Ratna Das (Senior Library Information Assistant, Central Library, IIT Delhi)
This analysis was generated by skim (skim.plus), an AI-powered content analysis platform by Credible AI. Scores and classifications represent the platform's AI-generated assessment and should be considered alongside other sources.