How Humans Succeed While Failing to Communicate (Invited Talk)

Author Jang F. M. Graat

Thumbnail PDF


  • Filesize: 241 kB
  • 8 pages

Document Identifiers

Author Details

Jang F. M. Graat
  • Smart Information Design, Amsterdam, The Netherlands

Cite AsGet BibTex

Jang F. M. Graat. How Humans Succeed While Failing to Communicate (Invited Talk). In 9th Symposium on Languages, Applications and Technologies (SLATE 2020). Open Access Series in Informatics (OASIcs), Volume 83, pp. 1:1-1:8, Schloss Dagstuhl – Leibniz-Zentrum für Informatik (2020)


Humans communication is full of errors, but this does not prevent us from achieving common goals. Most of the constant flow of misunderstanding is not even noticed. We think we understand each other - until the moment when something stops making sense. If the common goal is important enough to us, we will ask for explanation: "What do you mean with that?" After receiving more information, we backtrack and correct the reconstruction of meaning in the conversation. Humans are fault-tolerant communicators because nothing is ever fixed: meanings are always temporary constructions and can be changed at any time, when they appear to be outdated or misconstrued. We can "change out minds" about something and we do not need to have perfect knowledge to survive. Those humans who strive for certainty in their lives are often called control freaks, i.e. unnatural exceptions. Computers live on the other side of the certainty spectrum: a 1 is a 1 and a 0 will always be a 0. Even if the programmer tells the computer - via string of 1s and 0s - to do something really stupid, the computer will simply do it. No room for error, as every decision is precisely defined. Of course, layers of automated translations between our modern programming languages and the processor’s basic instructions may introduce errors, but those are never called "misunderstandings". Instead, they are called bugs and they can be found and corrected, after which everything will be in full control again. The task of making a computer understand human communication therefore seems to be the hardest thing to do. It requires mechanisms of uncertainty, backtracking and feedback loops, each of which need to take into account that no meaning is ever fixed. In the end, the goal of understanding human communication remains subordinate to the goal of putting the communication partners in a position where they can achieve common goals. Whether the communication partners are both human, both computer or a mix of the two does not really change that. Communication of any type is always a means, never a goal.

Subject Classification

ACM Subject Classification
  • Human-centered computing → Interaction paradigms
  • Natural Language processing
  • Communication


  • Access Statistics
  • Total Accesses (updated on a weekly basis)
    PDF Downloads