online chat
Enable an AI agent to recognise an online chat, assess its conversational state, and determine which participation, messaging and moderation actions are permitted.
Research draft, second pass
A second pass drafted this model: the structure a model of this thing needs, and what is known about it in the world. The line under this one says how the second half was obtained - researched against sources, or recalled without web access, in which case nothing here was read anywhere and every claim is a lead to verify. Unreviewed either way.
Researched by: Codex + Grok
Purpose and description
Enable an AI agent to recognise an online chat, assess its conversational state, and determine which participation, messaging and moderation actions are permitted.
Online chat is a computer-mediated communication mode in which two or more parties exchange discrete, typically typed messages over a packet network with an expectation of near-real-time delivery and interleaved turn-taking inside a live session (a scrolling transcript, room, or thread), as distinct from store-and-forward mail and from continuous audio or video telephony.
It can be Join, leave or invite participants when the chat's access rules and the agent's authority permit it; Read and summarise accessible conversational context while preserving uncertainty about missing history; Compose and send an authorised message to the correct chat, recipients and thread; Ask for clarification, answer an outstanding turn or hand the conversation to another participant; Edit, retract or moderate messages where supported and authorised; Mute, pause, close or reopen participation while distinguishing personal settings from shared conversation state.
Distinguishing features
Participants contribute messages to a shared conversational context; a one-way notification feed alone does not establish an online chat.
Messages function as conversational turns that can receive responses within the same context; publication beneath a post alone does not establish chat.
An addressable conversational space connects successive messages; unrelated direct notifications do not become one chat merely because they reach the same recipient.
Human participation is not required in every turn: a person-to-agent exchange may qualify when both sides participate in the same conversational context.
Simultaneous presence is not required: delayed replies may continue the same chat when its identity and conversational context persist.
Scope
+ The identity and boundaries of a chat, including its relationship to rooms, threads and sessions
+ Participants, conversational roles, presence and rights to participate
+ Message exchanges, reply relationships and delivery or visibility states
+ Conversation progress, unanswered requests, interruptions and closure
+ Chat-specific access, moderation and agent participation constraints
- The hosting application's infrastructure, commercial offering and service operations
- Participants' complete identity, account and contact records
- The internal structure of documents, images or other attachments exchanged in chat
- Work items, transactions or commitments created through conversation
- Email correspondence, public comment systems and audio or video meetings as independently modelled activities
Characteristics
- Conversational topology
- One-to-one, group, open room, or unresolved Determines whether addressing one participant also exposes a message to others.
- Participation mode
- Human-only, human-agent, agent-only, or unknown Helps interpret conversational roles and determine whether automated participation needs disclosure.
- Participation authority
- Participant-to-chat permissions for reading, posting, inviting, editing and moderating Separates the ability to access a chat from authority to act within it.
- Observed presence
- Per participant: reported available, away, offline, or unknown, with observation time Supports response expectations without equating a presence indicator with attention.
- Latest observed activity age
- Seconds since the latest observed conversational message, measured at a stated time Helps distinguish an active exchange from a dormant conversation without declaring it closed.
- Reply dependency
- Message-to-message or message-to-thread relationship; explicit, inferred, or unknown Prevents an agent from answering the wrong turn in an interleaved conversation.
- Message transport state
- Per message and relevant recipient: pending, service-accepted, delivery-confirmed, failed, or unknown Prevents a local send attempt from being treated as confirmed delivery.
- Conversation disposition
- Ongoing, awaiting response, paused, explicitly closed, or uncertain Supports a decision to reply, wait, resume or leave the exchange closed.
- History accessibility
- Complete within a stated boundary, partial, unavailable, or unknown Limits claims about what participants previously said or agreed.
Also called
Where this came from
wikidata · CC0 1.0
Drafted structure
Bundle to layer to finding to question, as the second pass will find it: 6 bundles · 11 layers · 18 findings · 36 questions.
Conversation boundaries Establishes which conversational activity is being modelled and how its context persists.
An agent must distinguish continuing a chat from entering a different audience or starting a separate exchange.
Chat identity
Identifies the shared message space and its relationship to conversational subdivisions.
Addressable conversation
Record the chat's addressable identity and whether rooms, threads or sessions are parts of it or separately tracked conversations.
- What identifies this chat independently of its editable title or participant display names? definition
- Does entering a thread continue this chat or establish a separately bounded conversational activity? boundary
Context continuity
Describes which earlier turns belong to the conversation and remain accessible.
Available conversation history
Record the accessible history boundary, known gaps and any evidence that a resumed exchange continues an earlier chat.
- Which time range and threads are represented in the history available to the agent? measurement
- What evidence connects this resumed exchange to the earlier conversation? provenance
- What missing context must be requested before relying on an earlier decision? action
Participants and audience Tracks who participates, who can observe and whom a message addresses.
Chat membership, actual attention and intended recipients can differ, affecting both interpretation and disclosure.
Membership and roles
Separates conversational identities from membership and operational authority.
Participant standing
Record participant references, known human or agent status, membership changes and chat-specific roles.
- Which identities are members, guests, agents or moderators in this chat? definition
- What evidence establishes each participant's current role and membership? provenance
Addressing and attention
Distinguishes the visible audience, intended addressees and observed presence.
Recipient and presence context
Record who can see a turn, who is being addressed and what presence indicators actually report.
- Who can see this message, including observers who are not its named addressees? boundary
- What presence or typing signal was observed, and when? measurement
- Should the agent reply to the whole chat, use a thread or seek an authorised private exchange? action
Message exchange Represents conversational turns, their relationships and their observable transmission history.
An agent must know what a turn responds to and whether its own contribution was sent, changed or received.
Turn relationships
Preserves reply structure when messages arrive concurrently or refer to older turns.
Conversational turn context
Record the author, conversational placement, reply target and relevant attachments or quoted material for a turn.
- Which earlier message or conversational branch does this turn respond to? definition
- Is the reply relationship explicit in the chat or inferred from message content? provenance
- Does intervening conversation change whether the proposed reply is still relevant? action
Message lifecycle
Distinguishes submission, delivery, reported reading and subsequent mutation.
Message state evidence
Record observable send, delivery, read, edit and deletion events without treating them as proof of understanding or agreement.
- What evidence supports the message's current delivery or reported read state for each relevant recipient? provenance
- Which version of an edited message is visible, and is earlier content accessible? boundary
- Would retrying a send risk duplicating a message whose delivery remains uncertain? action
Conversational progress Tracks what the exchange is trying to resolve and which conversational move is appropriate next.
Recent activity alone does not reveal whether an answer, clarification, handoff or silence is appropriate.
Pending turns
Identifies unanswered requests and uncertainty about who should respond.
Response expectations
Record outstanding questions, their intended respondents and whether later turns answered, withdrew or superseded them.
- Which question or request remains unanswered, and who is expected to respond? definition
- Which messages support treating that request as answered, withdrawn or superseded? provenance
- Should the agent answer now, clarify the request or wait for another participant? action
Pause, handoff and closure
Distinguishes inactivity from agreed pauses, transferred responsibility and explicit endings.
Conversation disposition evidence
Record evidence for the exchange's current disposition and any acknowledged transfer of conversational responsibility.
- Was the conversation explicitly paused or closed, or has it merely become inactive? boundary
- Has another participant acknowledged taking over the conversation? provenance
- What event or authorisation would justify resuming the exchange? action
Chat action controls Captures the rules governing access, automated participation and intervention in this chat.
A technically available chat action may exceed the agent's authority or change visibility for other participants.
Access and agent authority
Connects chat capabilities to actual permission and the agent's delegated role.
Authorised participation
Record what the agent may read, disclose and send, whose behalf it acts on and any applicable automation disclosure requirements.
- Whose authority permits the agent to participate, and what messaging actions does it cover? provenance
- Which accessible content may be quoted or forwarded to this chat's audience? boundary
- Must the agent disclose its automated role before contributing here? action
Moderation and visibility effects
Describes the scope and consequences of chat-specific intervention.
Intervention scope
Record available moderation actions, their authority requirements and whether their effects are personal, shared or uncertain.
- Does muting, blocking, deleting or closing affect only this participant's view or the shared chat? boundary
- What source establishes who may remove messages, restrict participants or close posting? provenance
- Which authorised intervention addresses the observed issue, and what visible content or participation will it change? action
Evidence and external alignment What the world already says about this thing, gathered so the model can be checked against it.
A model that cannot be lined up against existing standards, identifiers and practice cannot be adopted by anyone who already uses them.
Reported evidence
Findings from the breadth pass, kept separate from the structural claims.
Kinds and varieties
Reported by the breadth pass; each item needs checking against its source before it becomes normative.
- One-to-one instant messaging (presence-aware direct messages)
- Multi-user rooms, channels, and group chats
- Internet Relay Chat (IRC) and other federated text-chat networks
- Customer-facing live web chat (on-site support widgets and contact-centre queues)
- Workplace team messaging (persistent channels plus threads)
- Live-stream audience chat (overlay beside a broadcast)
- In-game chat (team, party, and all-chat)
- Mobile consumer messengers, including end-to-end-encrypted apps
- Which of these kinds and varieties hold for the sense of online chat this model covers, and on what evidence? provenance
Identifiers and schemes
Reported by the breadth pass; each item needs checking against its source before it becomes normative.
- Wikidata - Q1634122 - Item commonly used for online chat as a communication mode; instant messaging is the related item Q58199.
- XMPP JID (RFC 6122 / RFC 7622) - [localpart@]domainpart[/resource] - Federated address of a chat user, room, or resource - not a name for the class itself, but the dominant identifier of chat endpoints on XMPP.
- IRC naming - #channel or ##channel on a named network; nick up to protocol limits - Channel and nickname identifiers on a given IRC network; networks themselves are named out of band (e.g. Libera.Chat).
- Matrix identifiers - @user:server ; !roomid:server ; #alias:server - User, room, and alias IDs on Matrix homeservers.
- Which of these identifiers and schemes hold for the sense of online chat this model covers, and on what evidence? provenance
Standards and regulation
Reported by the breadth pass; each item needs checking against its source before it becomes normative.
- RFC 1459 and RFC 2810-2813 (IRC architecture and client protocol) - IETF
- RFC 2778 / RFC 2779 (presence and instant-messaging model and requirements) - IETF
- RFC 6120 / RFC 6121 (XMPP Core and Instant Messaging) - IETF
- RFC 3428 (SIP extension for instant messaging, the MESSAGE method) - IETF
- RFC 9420 (Messaging Layer Security, MLS) - IETF
- RCS Universal Profile (rich chat over mobile networks) - GSMA
- Matrix Client-Server API - The Matrix.org Foundation
- Regulation (EU) 2016/679 (GDPR), for personal data in chat logs, accounts, and metadata - European Union
- Regulation (EU) 2022/2065 (Digital Services Act), for hosting and moderation duties of chat platforms - European Union
- Children's Online Privacy Protection Act (COPPA), 15 U.S.C. §§ 6501-6506 - United States
- Online Safety Act 2023 - United Kingdom
- Which of these standards and regulation hold for the sense of online chat this model covers, and on what evidence? provenance
Real-world use
Reported by the breadth pass; each item needs checking against its source before it becomes normative.
- Retail and SaaS websites embed a live-chat widget so a visitor can message a human agent or bot during a purchase or support incident; the transcript is often filed as a ticket.
- Workplace teams coordinate in persistent channels (Slack, Microsoft Teams, Mattermost) beside issue trackers and documents.
- Consumer messengers (WhatsApp, WeChat, iMessage, Telegram, Signal) are the default interpersonal channel in many countries, carrying 1:1 and group text plus voice notes, images, and payments.
- Open-source, operator, and hobbyist communities still run public IRC, Matrix, or Discord rooms for development and support.
- Twitch, YouTube Live, and similar services show a public scrolling chat beside a live video stream.
- Multiplayer games expose team, party, and all-chat for coordination and social play, often with voice as a parallel channel.
- Contact-centre platforms queue web and in-app chat as one customer channel among voice, email, and social.
- Which of these real-world use hold for the sense of online chat this model covers, and on what evidence? provenance
Typical measurements
Reported by the breadth pass; each item needs checking against its source before it becomes normative.
- One-way message latency (send to first delivery) - 20-500 on well-provisioned IM; up to several seconds on congested mobile or heavily moderated rooms - ms
- Casual message body length - 1-500 typical; protocol caps include IRC 512 bytes per line including framing - characters (or bytes for protocol caps)
- Concurrent participants in one room or channel - 2 (DM); about 3-15 for small groups; 10^2-10^3 for large team channels; 10^4-10^5 for popular live-stream chats - persons
- User message rate in an active session - 0.1-10 - messages per minute
- Session or presence duration - minutes to several hours for a sitting; workplace channels are effectively always-on - min
- Server-side retention of transcripts - 0 (client-ephemeral or disappearing messages) to unbounded - d
- Which of these typical measurements hold for the sense of online chat this model covers, and on what evidence? provenance
Failure modes and hazards
Reported by the breadth pass; each item needs checking against its source before it becomes normative.
- Message loss, duplication, or reordering under partition, netsplit, or at-least-once queues.
- Nickname or account impersonation and session hijack (stolen cookies, SIM swap, credential stuffing).
- Flooding, spam, and denial of service against a room or the chat service.
- Harassment, doxxing, coordinated raids, and grooming; public and semi-public rooms are a known vector for child sexual exploitation material.
- Assumed ephemerality broken by logs, backups, search indexes, screenshots, and forwards.
- Traffic analysis and metadata leakage (who talked to whom, when) even when bodies are end-to-end encrypted.
- Phishing and social-engineering payloads delivered inside a trusted chat thread.
- Split-brain and history divergence on federated networks after netsplit or homeserver outage.
- Moderation failure: either under-blocking of abuse or over-blocking of legitimate speech, often with poor appeal paths.
- Which of these failure modes and hazards hold for the sense of online chat this model covers, and on what evidence? provenance
Regional variation
Reported by the breadth pass; each item needs checking against its source before it becomes normative.
- Default consumer app differs by market: WeChat in mainland China, LINE in Japan, Taiwan and Thailand, KakaoTalk in Korea, WhatsApp across much of Latin America, India, Africa and Europe, iMessage among US Apple users, Telegram strongly used in parts of the CIS, Iran and MENA.
- China requires real-name registration and extensive content controls on public chat; the EU emphasises GDPR/DSA duties; the US still leans on intermediary-liability rules (including CDA §230) plus COPPA for children.
- Ordinary language splits: English 'chat' vs 'IM' vs 'messenger'; many locales treat WhatsApp/WeChat as the generic word for the activity.
- IRC and XMPP remain more visible in Western free-software and operator culture; consumer markets in East Asia never treated IRC as the default.
- Which of these regional variation hold for the sense of online chat this model covers, and on what evidence? provenance
Neighbouring kinds and how to tell them apart
Reported by the breadth pass; each item needs checking against its source before it becomes normative.
- Email - Store-and-forward with no live session or presence requirement; acceptable latency is minutes to hours, not sub-second turn-taking.
- SMS / MMS / RCS as a telephony service - Addressed by MSISDN and carried on mobile-operator messaging infrastructure (even when the UI looks like chat); online chat in the Internet sense is addressed by accounts, JIDs, or room IDs on IP services.
- Internet forum / bulletin board / mailing list - Asynchronous, post-and-thread, not a live interleaved session; participants do not expect the other party to be present now.
- Voice or video call (PSTN, VoIP, conferencing) - A continuous audio or audiovisual stream is the primary channel; chat, if present, is an auxiliary data channel.
- Social-media comments and replies - Turns are anchored to a public post or object and are not a free-standing live session between the parties.
- Collaborative document comments - Object-anchored annotation of a shared artefact, not a standalone conversational session.
- Which of these neighbouring kinds and how to tell them apart hold for the sense of online chat this model covers, and on what evidence? provenance
Sources
- Internet Relay Chat Protocol (RFC 1459) - Canonical early Internet chat protocol: session model, channels, nicknames, and the 512-byte message ceiling that still shapes IRC-family practice.
- Instant Messaging / Presence Protocol Requirements (RFC 2779) - IETF requirements that separate instant messaging and presence from email: near-real-time delivery, identity, and the presence model used by later IM systems.
- Extensible Messaging and Presence Protocol (XMPP): Core (RFC 6120) - The IETF standard for federated XML messaging and presence, including JID addressing used as an identifier scheme for chat endpoints.
- Language and the Internet (2nd ed.) - Linguistic and CMC account of chatgroups versus email, the web, and virtual worlds; supports the definition of chat as synchronous typed interaction and several neighbour distinctions.
- Online chat - Overview of the public sense of the term, the main service kinds (IM, IRC, web chat, game chat), and common neighbouring concepts.
What the second pass must settle
- Does the registry intend online chat to denote a conversational activity, a persistent chat space, or both, and does an existing world model already own either concept?
- Where should the operational boundary fall between online chat, threaded public comments, email-like messaging and chat embedded in an audio or video meeting?
- Should threads and resumed sessions receive independent model identities, or remain subdivisions of a persistent chat?
- Which delivery, read, presence, editing and deletion signals can be interpreted consistently across supported services, and which require service-specific definitions?
- Should agent-only exchanges qualify as online chat under this registry entry, or belong to a neighbouring model of inter-agent communication?