Overview
On 5-6 August 2025, the All-domain Anomaly Resolution Office sponsored a two-day workshop at Associated Universities, Inc. (AUI) in the Washington, D.C. area devoted to a problem that sits beneath almost every modern UAP debate: how should large collections of human reports be preserved, compared and linked to technical evidence?
The meeting, titled “2025 UAP Workshop: Narrative Data, Infrastructures, and Analysis,” brought together 40 participants from government, academia and independent research organisations. It was organised through a collaboration involving AARO, AUI and Florida State University. AARO — 2025 UAP Workshop synthesis AUI — workshop report release
The workshop itself occurred six months before AARO publicly released its synthesis paper on 13 February 2026. That later publication has a separate UAPRAD deep dive. This page concerns the August 2025 event: why it was held, how participants were selected, what kind of discussion took place, what the structure reveals about AARO's developing relationship with outside researchers and what can—and cannot—be inferred from a government-sponsored meeting of this kind.
The event was deliberately small and invitation based. AARO's published invitation says organisers originally expected roughly 25-30 experts; the final group grew to 40. Potential attendees were sought for expertise in artificial intelligence and machine learning, UAP research and data, physical and natural sciences, information and data science, archives and records, analytical methods, cyberinfrastructure and computation, and human and social sciences. Invitees who declined could be replaced by candidates with similar backgrounds identified through research and recommendations. AARO — workshop paper, invitation and participant-selection sections
Participants were selected because organisers believed they could contribute to the problem. The meeting therefore records expert discussion among a chosen interdisciplinary group, rather than a consensus of the wider scientific community. Its institutional importance lies in bringing government analysts into the same small working environment as academic information scientists, archival specialists, AI researchers and representatives of civilian UAP organisations whose databases had historically existed outside the government's own reporting system. That development should be separated from participants’ individual conclusions about UAP origin.
Chronology
15–17 May 2024 — NSF-supported predecessor workshop.
ASTRO ACCEL convenes an interdisciplinary meeting on UAP science, public engagement and communication.
4 August 2025 — pre-workshop networking.
Participant-side accounts identify a contextual networking event before the formal AARO workshop.
5–6 August 2025 — AARO narrative-data workshop.
AARO, AUI and Florida State University convene 40 selected participants to consider UAP reporting, data infrastructure and analysis.
11 August 2025 — participant account published.
NUFORC publicly describes its involvement and the workshop’s discussions.
2 September 2025 — further participant reporting.
The Society for UAP Studies publishes a longer account of the symposium.
13 February 2026 — AARO synthesis released.
The office publishes its workshop paper, making the agenda, selection framework and recommendations more readily assessable.
Workshop context and participation
The evidence problem
AARO's later synthesis states that the meeting centred on the collection, organisation and interpretation of UAP reports, especially narrative material. The organisers sought to assess reporting systems and repositories, identify weaknesses in collection and access, explore analysis and pattern-recognition methods, build trust between sectors and develop stronger data infrastructure. AARO — workshop synthesis
Those aims do not assume that UAP are one physical phenomenon. A database of unusual aerial reports can include aircraft, balloons, satellites, meteors, sensor artefacts, hoaxes, misperceptions and genuinely difficult cases. Before any residual pattern can be evaluated, researchers need to know whether apparently similar reports were collected comparably: a cluster may reflect a physical pattern, but may also reflect media attention, reporting campaigns, observer density or differences between an airline-pilot database and anonymous public submissions. The August workshop addressed this foundational evidential problem rather than an origin hypothesis.
Narrative data
Human reports are often treated as the weakest form of UAP evidence, particularly when a narrative is asked to establish quantities it cannot reliably measure.
A witness can describe an object as moving “extremely fast” without knowing its distance. The report is then strong evidence for the witness's perception of rapid angular motion but weak evidence for a particular physical velocity.
The workshop did not argue that narrative evidence should replace sensors. It treated reports as information capable of locating and contextualising other evidence: a report may supply a date, approximate time, location, viewing direction, aircraft callsign, weather description, additional witnesses or the existence of a photograph, radar track or police record. These details can lead investigators toward independently checkable sources.
Narrative evidence also contains information that structured forms can accidentally destroy. A witness may describe an unusual sequence or uncertainty that does not fit a predetermined category; if a system permits only one shape, object count or speed estimate, the form itself may reshape the observation. The workshop therefore treated the task as preserving both human meaning and analytical structure, a central tension of the two days.
The 2024 predecessor
The 2025 meeting had a direct predecessor. From 15–17 May 2024, an interdisciplinary workshop titled “Unidentified Anomalous Phenomena (UAP): A Dialogue on Science, Public Engagement and Communication” was held in the Washington area under the National Science Foundation-supported ASTRO ACCEL programme. Its organisers included Gretchen Stahlman of Florida State University and Tim Spuck of Associated Universities, Inc. 2024 UAP Workshop Executive Summary — Zenodo Florida State University — June 2024 faculty brief
The earlier event focused more on communication, public engagement, stigma and the social conditions of UAP research; the 2025 workshop moved toward infrastructure. The connection should not obscure the funding distinction: the 2024 workshop was NSF-supported through ASTRO ACCEL, whereas the 2025 narrative-data workshop was sponsored by AARO. It is therefore not an NSF UAP research programme or evidence that NSF adopted AARO’s agenda. AUI describes the later meeting as building on the earlier work through a new AARO/AUI/FSU collaboration. AUI — February 2026 statement
Private participation
The 2025 meeting was not publicly advertised in advance, a decision that attracted attention after it became known. AARO’s synthesis gives a direct explanation: participant privacy was considered important, and an Institutional Review Board approval governed workshop data collection and security. Organisers sought a neutral environment in which people with differing beliefs and professional backgrounds could speak without immediate public attribution, so the committee did not publicise the workshop online beforehand. AARO — workshop paper
This was an invite-only private workshop, not evidence of a classified UAP briefing.
Participants were told not to photograph other attendees without consent and not to attribute statements without permission. Name badges could contain only a first name and no institutional affiliation. Participants could also leave sessions if they were uncomfortable with a topic.
Later reporting by DefenseScoop says attendees paid their own travel to the Washington area while AARO covered lodging and most meals, matching the invitation reproduced in the paper. DefenseScoop — AARO's 2025 invite-only workshop
Privacy measures can create an appearance of secrecy from outside. In this case, the published documentation identifies a methodological reason: enabling frank discussion where professional reputations or organisational relationships could be affected by public association with individual remarks. This may have supported candid participation while reducing transparency about who attended and how representative the group was.
Participant composition
AARO did not publish a complete roster, preventing a precise independent audit of disciplinary balance. The office’s statement that participants came from government, academia and independent research organisations across specified expertise areas remains the strongest official account. Several organisations later identified their own representatives publicly.
The National UFO Reporting Center said its chief technical officer Christian Stepien attended and described other represented organisations including the National UFO Historical Records Center, Scientific Coalition for UAP Studies, Galileo Project, UAPx, Society for UAP Studies and Archives of the Impossible. NUFORC — workshop participation
The Society for UAP Studies published a more detailed participant-side account, identifying Robert Powell of SCU, David Marler of the National UFO Historical Records Center, Christian Stepien of NUFORC, physicist Matthew Szydagis, archivist Amanda Focke, social scientist Ben Gibson and several SUAPS representatives. Society for UAP Studies — AARO and AUI Engage Academia These accounts improve visibility but do not constitute an official complete list.
They nevertheless indicate that the room included people who regarded UAP research as deserving sustained scientific attention and who represented organisations devoted to that purpose. This may have enriched practical knowledge of historical databases while shaping the range of assumptions represented; the available material does not permit the balance of sceptical, agnostic and extraordinary-origin views to be measured.
AARO and civilian organisations
AARO Director Jon Kosloski gave opening remarks. Participant-side reporting says he emphasised the office's focus on UAP as a problem of domain awareness: reducing gaps in the government's ability to detect, identify and attribute objects or phenomena operating in areas of national-security interest. NUFORC — workshop report Society for UAP Studies — symposium account
This framing separates the meeting from an origin-driven programme. An unusual report is operationally important when it reveals uncertainty about what is present in a monitored domain; a balloon, drone, satellite, sensor artefact or more unusual object can initially reveal that gap. The workshop’s emphasis on narrative information, standardisation and linkage follows from this role by helping investigators find sensor data, identify known activity and distinguish repeated reports from independent events.
Data infrastructure and analysis
Existing databases
The published agenda records an opening keynote on the importance of good UAP data. Participant accounts identify the speaker as Robert Powell, co-founder and board member of the Scientific Coalition for UAP Studies, who discussed the history and quality of existing databases, how archives were created, natural-language processing and automated scientific detection systems. Society for UAP Studies — symposium account
This was an appropriate opening theme because legacy UAP databases are often compared as though they contain equivalent observations. They do not: one may collect public submissions, another pilot reports, a military system may retain sensor metadata unavailable to civilians, and an archive may hold newspaper accounts or historical interviews. Even fields such as shape and duration can use different definitions. Analysis that overlooks those differences can produce patterns that reflect the databases rather than the sky.
The keynote therefore established a methodological premise that recurred throughout the workshop: before combining datasets, researchers need to understand how each dataset was made.
Integration and provenance
The first formal breakout mapped a data landscape that included historical files, current narratives, radar and imagery, flight, weather, astronomical and seismological records, social media and other contextual sources. A report of a light at a precise time and location can potentially be compared with ADS-B tracks, weather radar, satellite passes and fireball records, but the datasets do not naturally fit together: time zones differ, locations may be prose rather than coordinates, identities may be protected, military information can be classified, historical terminology varies and public platforms can restrict access.
Participants therefore favoured modest pilot-scale integration projects rather than an immediate universal merger. A unified database may be attractive, but poorly harmonised records can create false precision. Interoperability requires common definitions and explicit mappings between fields, not simply placing every record on one server.
Crosswalks and comparability
A database becomes analytically dangerous when fields appear equivalent but are not. One archive may assign “shape” from a fixed list, another allow free text and a third apply a researcher’s category after submission; merging them under a single shape field can conceal those differences. The same applies to witness type: “pilot” can mean a military fighter pilot, airline captain, private pilot or even a passenger if categories are poorly defined. Data dictionaries, crosswalks and metadata standards are therefore safeguards against invalid comparison rather than administrative detail.
Artificial intelligence received a dedicated panel because UAP archives contain large quantities of unstructured text. A language model can extract dates, locations, shape terms, witness types and candidate explanations faster than a person reading each report; semantic search and clustering can generate hypotheses about related records. The workshop did not treat those capabilities as a shortcut to determining what UAP are.
Participants instead identified hallucination, cultural bias and data contamination as central risks. Public UAP material contains duplicated stories, folklore, speculation, hoaxes and repeated reinterpretation, so a model can reproduce cultural expectations as though they were evidence. For example, repeated online references to “Tic Tacs” may make a system more likely to apply that label to a vague oblong description. Human-in-the-loop methods are valuable precisely because AI may propose classifications, but people must verify them against source records.
Duplicated records
One UAP event can generate a military report, a journalist’s interview, a civilian database entry, social-media reposts and later archival copies. An automated system may then count five dependent records as five independent observations, creating false corroboration and artificial geographic, temporal or morphological clusters. Provenance is therefore essential: every record should preserve where it came from and what other records it depends upon. A graph of source relationships may be more useful than a flat spreadsheet, and raw report counts should be treated cautiously across historical datasets.
The second breakout addressed a methodological tension: statistical analysis requires comparable variables, whereas aggressive reduction can remove what makes a narrative meaningful. The synthesis records support for both corpus-level methods—time and geographic clustering, keyword trends, correlation and graph analysis—and qualitative approaches including phenomenology and discourse analysis. This is an argument about preserving data, not an endorsement of every experiential claim in UAP literature.
A witness’s report that an object “responded” contains at least two layers: the sequence of observed movements and the interpretation that the movements were responsive. A well-designed system preserves both without confusing them. Storing observation and interpretation separately allows the former to be tested without discarding the witness’s subjective account.
Experiential reports
Participant accounts indicate that the meeting included archival, social-science and experiential perspectives alongside physical science and AI. Amanda Focke of Rice University’s Archives of the Impossible reportedly argued that witness narratives should not be reduced to “mere anecdotes” and that reports of communications, interactions or consciousness-related experiences should be retained as qualitative data. Society for UAP Studies — symposium account
Preserving such testimony is not the same as validating its interpretation. A claimed communication with non-human intelligence can be studied as a human report without presuming that the claimed intelligence exists. Such material may carry little weight in physical identification without independent corroboration, while remaining central to research on witness experience, reporting culture or psychological effects. The workshop’s approach is methodologically defensible only where these analytical layers remain distinct.
Reporting-system design
The third breakout considered a hypothetical online system containing 1,000 UAP reports stored as PDF files and asked how it could be cleaned, reorganised and improved. The exercise concerned reporting-form design rather than UAP ontology. Participants suggested free narrative or optional audio before structured fields proposed by AI and verified by the witness. Questions should address perceived properties—angular size, apparent motion, sound, shape and effects—rather than compel exact physical estimates, and should ask how size, speed or distance was estimated.
They also recommended approximate witness counts rather than a binary “mass sighting” field, automatic ingestion of image metadata where possible, back-end normalisation of dates and locations, and retention of unknown values rather than guesses required by a form. AARO — workshop synthesis These recommendations show how data quality can be damaged at collection: a poorly designed form can manufacture precision the witness never possessed.
Taxonomy and witness description
The workshop recommended a concise shape taxonomy—disk, sphere, triangle, cigar and other—while retaining free-text capability. Taxonomy aids comparison but can also influence observation: presenting familiar shape labels before a witness describes an ambiguous object may encourage selection of the closest cultural category, which can later be mistaken for an independently recurring pattern.
The proposed order reduces this risk by capturing a witness’s narrative first, then extracting or confirming structured categories. This is a useful principle for UAPRAD’s wider taxonomy work: a clear public-facing label can aid communication while the evidence layer retains the raw description and uncertainty.
Corroboration and witness status
Participants discussed how to triage reports for scarce investigative resources. Military personnel, pilots and police may have relevant observational training, but the workshop cautioned against automatically assigning excessive weight to particular professions. A professional witness may provide a stronger description, yet professional status does not supply missing range or eliminate expectation, memory error or sensor misunderstanding.
The workshop instead emphasised corroboration: a report becomes stronger where independent witnesses, radar, weather data, flight tracks or other records converge. This shifts evidential weight from the witness’s status toward the independent support available for the claim.
Institutional scope and public record
Privacy and ethics
The workshop’s structure mirrored the ethical problems it discussed. UAP reports can include names, home locations, medical or employment information, while government reports may contain classified sensor details. Public databases therefore cannot maximise openness without considering consequences. The workshop recommended reporter control over sensitive metadata and aggregated or de-identified releases where appropriate; its IRB-governed controls over attribution and photography applied the same principle to the meeting itself.
This approach privileges transparency about the content of the workshop over attribution of each remark. That is a limitation when assessing scientific consensus, but may support candid participation.
Research scope
The presence of scientists, officials and technical specialists does not make the meeting equivalent to peer-reviewed research. Participants did not publish independently replicated experimental data, formally establish a UAP pattern or assign physical explanations to a defined case set. The later synthesis draws on workshop notes, transcripts and organiser interpretation, as is normal for a workshop. Its outputs are best read as research-design recommendations and evidence of the infrastructure concerns shared by participants. The practical convergence is a strength; the unavailable full deliberative record and participant composition limit any estimate of how representative those views were.
The institutional relationship may be the workshop's most important feature. Civilian UFO organisations historically developed partly because government investigations were seen as inaccessible, dismissive or incomplete. NICAP, APRO, CUFOS, NUFORC and later groups built separate archives because government did not maintain one openly accessible system covering every public report. The August 2025 workshop brought some of that civilian ecosystem into direct discussion with AARO.
NUFORC publicly described the event as an opportunity for government, academic and private organisations to discuss interoperable systems. Its own representative presented the organisation's long-running reporting work and database. NUFORC — workshop report
This does not mean AARO endorsed the conclusions or reliability of those databases. It means the office recognised that useful information may exist outside government holdings and that researchers need methods to understand and compare it. That is a meaningful change from a model in which civilian archives and government databases operate almost entirely in parallel.
Institutional differences
The workshop did not erase disagreement between AARO and civilian researchers. The Society for UAP Studies account records participants expressing views that extend beyond AARO’s public analytical posture, including research on non-human intelligence, consciousness-related reports, alleged materials and possible links with nuclear activity. AARO’s stated mission remains narrower: to detect, identify, attribute and resolve anomalies relevant to domain awareness and national security.
Such differences need not prevent useful collaboration. The central question is whether common standards allow competing hypotheses to be tested against the same observations. A sceptical analyst and an extraordinary-hypothesis researcher should at least be able to agree on the timestamp, original image, provenance and sensor type; the workshop’s infrastructure emphasis creates the possibility of that shared evidential layer.
Public disclosure
Although the meeting was not advertised beforehand, its existence did not remain hidden until the 2026 AARO paper. NUFORC published an account on 11 August 2025, and the Society for UAP Studies published participant material in August followed by a longer account on 2 September 2025; AUI later listed the workshop among its 2025 Department of Defense work. NUFORC — 11 August 2025 Society for UAP Studies — 2 September 2025 AUI — Making Scientific Breakthroughs Possible in 2025
The event was private before and during the meeting, but not a programme whose existence remained concealed. Participant organisations disclosed their involvement within days and weeks, while AARO’s formal synthesis appeared later. The published record indicates that the delay reflected transcription, anonymisation, synthesis and prepublication review rather than classification of the workshop itself.
The 2026 synthesis
The workshop became substantially easier to assess when AARO published its 17-page synthesis. AARO’s records index lists the paper as added on 13 February 2026, after DoD prepublication review. AARO — UAP Records
The document discloses the invitation, selection framework, privacy rules, agenda, breakout prompts, discussion themes and recommendations, allowing participant publicity to be separated from the official record. It confirms that the workshop was selective, that 40 people attended, that AI had a dedicated panel, that qualitative and quantitative integration was discussed, and that the final breakout used a hypothetical 1,000-report system. It does not contain a hidden catalogue of solved cases, announce anomalous technology or claim that narrative clustering has identified a single physical phenomenon. Its output is a roadmap.
Implementation
The scientific value of a workshop depends on what follows. Its recommendations are concrete enough to assess: standard metadata templates, preservation of native media, cross-database mappings, human-supervised AI triage, links to air-traffic, weather and astronomical systems, better witness questions, de-identified public releases, performance benchmarks and further collaborative workshops.
By August 2026, AARO had continued expanding public case descriptions, sensor analysis and formal research products. Its FY2025 annual report describes multi-domain sensor and algorithmic prototyping, FAA reporting and improved modelling capabilities. AARO — FY2025 Consolidated Annual Report The connection is suggestive rather than evidence of direct implementation: AARO has not publicly shown that it adopted the complete roadmap, and no universal cross-sector metadata standard or interoperable public system has been demonstrated. The event remains an important design exercise whose institutional effects are still developing.
Evidence assessment
The August 2025 workshop is significant because it shifted attention from the interpretation of individual cases to how a research system should be built so that later conclusions are more reliable. That is a legitimate scientific and institutional problem. The meeting brought together disciplines that UAP research often keeps separate: physical scientists considering sensors, information scientists considering metadata and interoperability, archivists addressing lost provenance, social scientists separating observation from interpretation, AI specialists identifying both useful automation and new error pathways, and civilian organisations with practical knowledge of long-running public-report databases.
Its selective, invite-only design is both a strength and a limit. Privacy rules may have supported frank discussion, but they make independent evaluation of participant balance more difficult. The event should not be represented as a scientific consensus conference. Civilian UAP organisations may have contributed important knowledge of source material while also shaping the range of assumptions in the room; that selection context is relevant without being disqualifying.
The workshop's strongest methodological contribution is its attempt to retain both raw narrative and structured data. Reports need their original wording because unusual details and uncertainty can matter, but they also need metadata if they are to be compared at scale. That metadata must distinguish measured facts from estimates and interpretations. AI may help extract, cluster and triage information, but it cannot generate the underlying evidence or replace verification against source records.
The recommendations remain incomplete as an implementation programme. Interoperability between military, civilian and historical databases raises classification, privacy, legal and technical barriers that the meeting did not resolve; AI reliability requires measurable benchmarks; and a shared taxonomy can itself influence future reporting. The workshop is therefore best understood as a serious cross-sector design exercise for UAP evidence infrastructure, not as new evidence for any origin hypothesis. The August meeting concerns the organisational act; the February 2026 paper is its principal public documentary output.
The workshop also has a historical dimension. Civilian and government UFO research developed largely separate evidence systems: Project Blue Book held military files, NICAP and APRO built civilian archives, NUFORC accumulated public reports, government agencies retained classified sensor material, and academic engagement was usually episodic. The 2025 meeting brought parts of these cultures into one room to ask how their information could become mutually intelligible. Its longer-term value will depend on implementation: common schemas require adoption, cross-database analysis requires duplication and collection bias to be tracked, AI requires traceable outputs and measured error rates, and cross-sector collaboration requires enough unclassified evidence for independent analysis.
Source library
AARO — 2025 UAP Workshop: Narrative Data, Infrastructures, and Analysis
The controlling official record of the August event. The paper confirms the 5-6 August dates, AUI venue, AARO sponsorship, 40 participants, expertise-based selection, IRB-governed privacy measures, full agenda, breakout structure and the recommendations that emerged. The paper was published in February 2026, but it is the strongest primary source for reconstructing the 2025 workshop itself.
Read the official AARO workshop synthesis
AARO — UAP Records / Information Papers
The official publication index. It lists the workshop synthesis as added on 13 February 2026, allowing the August 2025 event to be distinguished cleanly from the later government publication event.
View AARO's UAP Records index
Associated Universities, Inc. — AARO workshop report release
Primary organiser statement describing the 2025 event as a collaboration between AARO, AUI and Florida State University and identifying rigorous collection, standardisation, analysis and responsible AI as central themes. It also confirms that the 2025 meeting developed from the earlier NSF-supported workshop.
Read the AUI statement
Associated Universities, Inc. — 2025 organisational review
AUI's own year-end summary states that it worked with the Department of Defense to design and implement a UAP workshop supporting a rigorous scientific framework and data-driven approach. Useful as independent institutional confirmation from the host organisation.
Read AUI's 2025 review
DefenseScoop — invite-only workshop reporting, 16 March 2026
Independent defence-technology reporting based on the published paper and Pentagon comment. Confirms the private/invite-only structure, attendee travel and lodging arrangements and AARO's statement that it hoped to convene future interdisciplinary workshops as needed.
Read the DefenseScoop report
2024 UAP Workshop Executive Summary — Zenodo
Primary published record of the NSF-supported predecessor meeting, Unidentified Anomalous Phenomena (UAP): A Dialogue on Science, Public Engagement and Communication, held in May 2024. Useful for distinguishing the earlier science-communication emphasis and funding source from the AARO-sponsored 2025 data workshop.
Read the 2024 executive summary
Florida State University — June 2024 faculty brief
Official FSU record confirming Gretchen Stahlman's role co-leading the 2024 NSF-funded UAP workshop. Relevant to the continuity of organising expertise between the 2024 and 2025 meetings.
Read the FSU faculty brief
National UFO Reporting Center — workshop participation, 11 August 2025
Participant-organisation account published days after the meeting. NUFORC identifies Christian Stepien as its representative, names several other civilian research organisations represented and describes the keynote, panels, breakout discussions and practical data exercise. It is particularly useful for establishing that the event's existence became public shortly after it occurred.
Read NUFORC's workshop account
Society for UAP Studies — AARO and AUI Engage Academia in UAP Symposium, 2 September 2025
Detailed participant-side report identifying a number of speakers and attendees and recording perspectives that do not appear by name in AARO's anonymised synthesis. It is valuable for reconstructing the range of views represented, while remaining an interested participant source rather than an official complete attendee list.
Read the Society for UAP Studies account
AARO — Fiscal Year 2025 Consolidated Annual Report
Later official context for judging possible implementation. The report describes continued work on sensor and algorithmic prototypes, integration of FAA reporting and improved analytical capabilities, while not establishing that the complete workshop interoperability roadmap has been implemented.
Read the FY2025 Annual Report