Common Knowledge in Email Exchanges

We consider a framework in which a group of agents communicates by means of emails, with the possibility of replies, forwards and blind carbon copies (BCC). We study the epistemic consequences of such email exchanges by introducing an appropriate epistemic language and semantics. This allows us to find out what agents learn from the emails they receive and to determine when a group of agents acquires common knowledge of the fact that an email was sent. We also show that in our framework from the epistemic point of view the BCC feature of emails cannot be simulated using messages without BCC recipients.

💡 Research Summary

**
The paper presents a rigorous formal model of email communication that captures the essential features of modern email systems: sending, replying, forwarding, and the blind carbon copy (BCC) mechanism. The authors begin by defining a finite set of agents and a set of abstract notes that represent the informational content of messages. A basic message is either a simple send operation s(i,l,G) (agent i sends note l to a non‑empty group G) or a forward operation f(i,l·m,G) (agent i forwards an existing message m, appending a new note l, to group G). This captures both replies and reply‑all by appropriate choices of G.

To model BCC, the notion of an email m^B is introduced, where B is a (possibly empty) set of BCC recipients. Regular recipients receive the message but are unaware of B; BCC recipients receive the same message but only know that they belong to the hidden set, not its composition. Crucially, when an email is forwarded, the BCC list is omitted, mirroring real‑world email clients. The authors also assume that an email’s sender is the only party who knows the full BCC list, and that no agent is simultaneously a regular and a BCC recipient for the same email.

The dynamic aspect of email exchanges is captured by the concept of a legal state. A state consists of a finite set E of sent emails and, for each agent, the set of notes currently known. A partial order ≺ on E must exist that respects three constraints: (L.1) a forward can only occur after the original email and the forwarder must have been a participant (sender, regular recipient, or BCC) in the original; (L.2) an agent may send a note only if it already knows that note, which must have appeared in some earlier email; (L.3) analogous conditions hold for BCC handling. This ordering guarantees causality without requiring explicit timestamps, allowing the analysis of knowledge irrespective of the exact chronological sequence.

The epistemic language introduced contains two modalities: K_i φ (agent i knows φ) and C_G φ (group G has common knowledge of φ). The semantics are defined relative to a legal state, taking into account the uncertainty about BCC membership. For each message m, the factual information FI(m) is the set of all notes appearing anywhere in m or its forwarded ancestors. Regular recipients acquire the entire FI(m), while BCC recipients acquire only the notes they already know plus any newly appended notes, but they cannot infer the existence of other BCC participants.

The central technical contribution is a characterization theorem for when a group acquires common knowledge that a particular email was sent. The theorem states that a group G attains common knowledge of the fact “email e was sent” if and only if (i) G includes every regular and BCC recipient of e, (ii) the email e is preceded (in the partial order) by all emails that convey the necessary factual information, and (iii) every member of G can trace a chain of forwards/replies that links them to the original sender. This result elegantly captures the nesting of common knowledge that arises from successive forwards (e.g., C_{A1} C_{A2} … C_{Ak} m) and shows how BCC creates additional layers: BCC recipients share a separate common‑knowledge sub‑group about the fact that the regular group knows the message.

A striking finding is the non‑simulability of BCC. The authors prove that any system that omits BCC and attempts to reproduce the same observable message flow cannot replicate the epistemic effects of true BCC. Without BCC, agents can always infer the full recipient list, eliminating the hidden‑knowledge component that is essential for certain common‑knowledge configurations. Thus BCC is not merely a cosmetic feature but a fundamentally new epistemic primitive.

The paper concludes by linking legal states to “properly terminating email exchanges,” defined as states where no further forwards are possible. In such terminating states, the common‑knowledge characterization applies cleanly, offering a clear picture of the final distribution of knowledge after a conversation thread ends.

Overall, the work provides a comprehensive logical framework for reasoning about knowledge dynamics in email communication, highlights the subtle epistemic role of forwarding and BCC, and establishes rigorous conditions under which groups achieve common knowledge. It bridges distributed systems theory, dynamic epistemic logic, and practical considerations of electronic mail, opening avenues for further research on privacy, security, and coordination protocols in asynchronous communication environments.