2.10Communication Failures
Communication can fail even when a channel exists and every agent is trained successfully.
In this section you will
- Distinguish missing information from irrelevant information
- Distinguish a lost message from an ambiguous one
- Recognize a protocol mismatch between agents
- Match each symptom to the intervention that can actually repair it
Missing Information
Section titled “Missing Information”The sender holds something useful and does not transmit it.
Agent 1 can see the order ticket and never mentions the dish. The information exists in the system, the channel exists, and the two are not connected.
How to spot it: compare each agent’s private information against what it sends. Anything the receiver could act on and never hears is a candidate.
Where it comes from: usually a communication policy that was never pressured to find the useful message, either because was high enough to discourage sending anything, or because the bootstrapping problem from Learning Communication Protocols never resolved.
Irrelevant Information
Section titled “Irrelevant Information”Messages are sent and change no decision.
Agent 1 reports the pot temperature, which agent 2 can already see. Or reports the chopping board, which agent 2 cannot use. Bandwidth is spent, nothing changes.
How to spot it: ablate. Replace a message with and see whether performance moves. If it does not, that message was not doing anything.
Why it matters even when it seems harmless: it costs every step, it makes the protocol unreadable by burying the useful part, and under a narrower channel it displaces something that mattered.
Message Loss
Section titled “Message Loss”The protocol assumes reliable delivery and messages disappear.
Communication Constraints covered the mechanism. The sharp version is that a lost message becomes , which is a legitimate message meaning “say nothing”, so a protocol in which silence carries meaning misreads loss as content.
How to spot it: sweep and look at the shape of the decline. A protocol robust to loss degrades gently. One that depends on silence falls off a cliff.
Message Ambiguity
Section titled “Message Ambiguity”The message arrives and does not narrow the receiver’s options enough.
Agent 1 sends 1, meaning “help needed”. Agent 2 does not know whether to
fetch an ingredient, take over the stove, or start plating. It has been told
that something is wrong and not what.
How to spot it: look for messages after which the receiver’s action distribution stays wide. If receiving a message leaves the receiver as uncertain as before, it did not resolve anything.
Where it comes from: most often a capacity limit. With two messages available, several situations must share one symbol, and the receiver has to disambiguate from context. That is a design choice with a known cost, not necessarily a bug.
Protocol Mismatch
Section titled “Protocol Mismatch”Two individually successful agents interpret the same message differently.
Agent A learned 2 means bring an ingredient. Agent D learned 2 means
serve the dish. Both scored highly with their own partners. Paired, they
fail, and no amount of retraining either one alone identifies the fault.
How to spot it: cross-play. Pair policies that were not trained together and compare against their within-team scores. A large drop is the signature. the Adapt chapter makes this a measurement discipline.
Why it has no single-agent analogue: there is no such thing as a convention held by one agent. The failure lives in the relationship, so it is invisible to any evaluation that tests agents only with the partners they trained with.
Failure Diagnosis
Section titled “Failure Diagnosis”| Failure | What is wrong | First thing to try |
|---|---|---|
| Missing information | useful facts never sent | check private information against messages sent |
| Irrelevant information | messages change no decision | ablate messages and watch performance |
| Message loss | protocol assumes delivery | sweep |
| Ambiguity | message does not narrow options | check the receiver’s action spread after a message |
| Protocol mismatch | nothing, in either agent | cross-play against unfamiliar partners |
Knowledge check
Correct.
Not quite.
Protocol mismatch. Both protocols work internally and assign symbols differently, so nothing is wrong with either agent.
Correct, and the signature is the gap between within-team and cross-team performance. The other four failures would show up within the original team as well; this one is invisible until you pair across runs.
Ambiguity. The messages are not specific enough for the new partner.
Ambiguity would also hurt the original pairing, since the same messages carry the same amount of information regardless of who receives them. The 94% rules it out.
Message loss on the new channel.
The channel did not change; only the partner did. Loss would degrade both pairings.
Irrelevant information. The messages were never carrying anything useful.
If the messages carried nothing, removing them would not hurt, and the cross-play score would be closer to the within-team score. A large drop on repairing is evidence the messages mattered a great deal, to the partner that understood them.
Explanation
The diagnostic habit worth building: a failure that appears only on repairing is about the relationship, not about either agent.
Communication Failures Summary
Section titled “Communication Failures Summary”- Missing information: useful private facts are never sent.
- Irrelevant information: messages arrive and change no decision. Ablate to detect.
- Message loss: the protocol assumed delivery. Worst where silence carries meaning.
- Ambiguity: the message does not narrow the receiver’s options, usually a consequence of limited capacity.
- Protocol mismatch: two working protocols disagree. Nothing is wrong with either agent, the failure lives in the relationship, and only cross-play reveals it.