I'm Alex Kearney, a PhD student studying Computer Science at the University of Alberta. I focus on Artificial Intelligence and Epistemology.


Here are collected messy notes from Short's theory of signs. Bits, pieces, and excerpts that I found interesting.

Notes from Short 2007

  • the key insight of these discussions is the need for generality. Generality means thinking in the general sense. Grey is not a single perceptual experience, but rather a collection of classes and instances.

  • Saussere: linguistic rules determine language and speech; study on the former depends on an a priori study of the latter. This inspired structuralism and actions of people, & thus the actions and history of society depend on an abstract representation which individuals are unaware of.

It was suggested that semiology could be carried over from language to all social behaviour and signs in the broadest sense. This is problematic, as it does not provide a satisfying description of how thought and language manage to be about the world.

Peirce adds an interpretant: the response a sign elicits. A sign, an object, and an interpretant are only taken to be one of three in relation to the other.

I guess an important thing to bring up is that signs---through the process of interpretation---give rise to something beyond the impression.

Each category has to justify itself by induction, giving it a limited and approximate validity (CP 1.301)

Really foreswearing the a priori necessity.

The three flaws of Peirce's semiotic of 1868:

  1. Thought lacks objects not constituted by thinking

  2. A thought-sign being interpreted in a subsequent sign suggests infinite semiosis

  3. If significance depends on interpretants--actual or potential--being signs, then we have no non-circular account of significance.

Peirce fixes infinite semiosis by creating indices.

Icon signification outside of resemblance without thought.

Icons derive their meaning from the feeling evoked by resemblance, thus, non-conceptual creatures can be sign interpreters.

What a sign signifies is from an a priori relation independent fo the sign actually being interpreted. Is then a type of possible interpretant grounded in interpretability.

To make this line of argumentation, two additions are required:

  1. Subjunctive conditionals and counter-factuals represent realities irreducible to actual occurrences.

Subjunctive conditional: A form of counter-factual. E.g., "Bob died yesterday. If he had died tomorrow, he would have been 98 years old."

  1. If an interpretation can be grounded, it must be with respect to a purpose, for a ground is something with respect to a grounding (final causation).

Final Causation: interpretation is purposeful, so interpretation must be relative to a possible action. Interpretability must be real and independent of it actually being actualised: reality of what was, is or could be: potentiality.

Grounding: in metaphysics one thing grounds another, if one thing accounts for the other.

The final problem is infinite semiosis:

1887 pragmatic maxim (w3: 226)

Consider what effects which might conceivably have practical bearings, we conceive of the object of our consideration to have, then our conception of these effects is the whole of our conception of the object.


  • Conception is a function of knowledge
  • Meaning is inexhaustible
  • Explication is never complete

It is only through purposeful action---even if only a potential action---that worlds and thoughts relate to a world beyond themselves.

More on belief and habit in (w3:247, cf 512)

A stone, a star, or a person is continuous spatio-temporally: undivided though divisible at any number of points... A church, a club, a crowd is never simply the members it has - pg 80

Clearly, when we build systems which perform classification and clustering, we are identifying generalities and continuities in some sense. There are an assortment of examples and kinds which through some form of training are identified. There is some continuity of perceptions which give rise to "cat" , or 'cluster 0'; however, many times they


There is a contradiction between teleology and behviourism and reductive theories. This can be found in mechanistic arguments.

We omit such a clause, as many scientists have thought that subsumption under law suffices to explain the phenomena non-telelogcially.

Fixation of Belief => How to Make Our Ideas Clear

Different minds may set out with the most antagonist views, but, the progress of investigation carries them by a force outside themselves [Peirce's Nominalism] to on and the same conclusion [his realism]... the opinion which is ultimately fated to be agreed upon.


Truth isn't clear when we have more boundaries: china isn't the border, or the people, or the govt... Static sausserean approaches are limited in these fuzzy ways. Behaviour, experience, agent => Peirce.

When we add the interpretant, we add context, interaction, something beyond original sensation.

Concept that it relates to in a meaningful way outside of stimulus.

Only reason that it works in supervised learning is that it is giving.