Search This Blog

Tuesday 18 September 2018

Peirce's Symbolic Signs and the Symbol Grounding Problem (For my WSU Tutorial scholars)


In Charles Sanders Peirce's semiotic system, the symbolic form of signs is distinguished from the iconic and indexical forms. Indexical signs have meaning via causal pathways and due to their causes, which they indicate: smoke means fire, for example. Iconic signs have some representational element (often pictographic) that makes some part of their semantic content obvious, like emojis and some kinds of hieroglyphics, for example.

The signs used in latinate, Germanic, Francophone, Arabic, and Cyrillic languages are generally symbolic. These are atomic languages in which there are word-symbol signs comprised of letter symbols that are both based upon a formal system that is conventional: constructed using conventions and rules. There is no way of discerning the meaning of the word signs without a dictionary that maps meanings or semantic content to the sign.
“Usually, the symbols constituting a symbolic system neither resemble nor are causally linked to their corresponding meanings. They are merely part of a formal, notational convention agreed upon by its users. One may then wonder whether an AA (or indeed a population of them) may ever be able to develop an autonomous,semantic capacity to connect its symbols with the environment in which the AA is embedded interactively. This is the SGP.”

In the above quotation, an AA is an artificial agent. Think of it as a very smart AI robot that has to figure out what the symbols in a language mean without any dictionary. If the symbols were pictographic like hieroglyphics (iconic) then if the robot knows about what people with buckets and spears look like, it has a way of decoding the meaning of the symbols. In the case of indexicals, the AA simply learns the causal rules and figures out what cause is being indicated – probably using physics and basic deduction. In the case of formal non-pictographic symbols like we use in English (words comprised of letters), the AA robot has little chance of figuring out their meaning without a dictionary. Put simply, there is no way to tell – if one doesn’t know any English and has no dictionary and no one to translate – what the meaning of any given English word is.
  
An indexical sign is one that indicates its cause. C.S. Peirce’s famous example is a weathercock (this is a metal fin mounted on a bearing and placed on a roof. It often has the outline of a fowl or cock. It has the same purpose as a windsock on an airfield.) The weathercock indicates which way the wind is blowing. Peirce called it a dicent sinsign. Arguably the AA would not have too much trouble with determining the meaning of such signs if it had a way of understanding cause and effect.

Image result for weathercock

An iconic sign depicts or represents what it signifies directly. Icons and avatars in games and on the internet. Emoticons are another example. If you have a rough idea what a smile is and what it means then you won’t have much trouble with this sign: 😊. With an ability to associate visual patterns, the AA would have a good chance of determining the meaning of such signs. 

A symbolic sign is exemplified by all of the words in all of these sentences. If you did not know the language and had no dictionary, you’d be hard pressed to know the meaning of the symbolic signs of English (or German, or Italian, or French, or Cyrillic-derived Russian, and so on.) This is what people mean when they say ‘It’s all Greek to me!’. The AA would have trouble with Peirce's symbolic signs for precisely this reason. There is no indication or hint embodied in the sign itself to tell the AA what the meaning of the sign might be. The semantic and semiotic system is hard (probably impossble) to adduce from 'scratch' even with lots of text.