Custom Search

Saturday, December 27, 2014

ACT (Adaptive Control of Thought), ACT*

ACT (Adaptive Control of Thought)
ACT (Adaptive Control of Thought) is a cognitive architecture based on the assumption of a unified theory of mind. The goal of this cognitive theory is to explain how human cognition works and what the structures and processes of human memory, thinking, problem solving, and language are. The core of ACT is a production system with a pattern matcher that works on memory and perceptual-motor modules via buffers. The current version of adaptive control of thought (ACT-R) is based on the principle of rationality of the human mind. Simulations with ACT-R allow for predicting typical measures in psychological experiments like latency (time to perform a task), accuracy (correct vs. false responses), and neurological data (e.g., FMRI-data).

ACT* Adaptive Control of Thought Model

Another network model, Anderson's (1990) ACT* (Adaptive Control of Thought) attempts to take into account all of our human cognition such as language, learning, decision-making and so on. It is not surprising then that it is a complex model which can be difficult to explain. ACT* distinguishes three types of memory structures: declarative knowledge, procedural knowledge and working memory. Declarative knowledge is knowledge about facts and things. Comparatively, procedural knowledge is knowledge about how to perform actions. Working memory is the part of declarative knowledge that is currently working as explained in Baddeley's approach.
According to Anderson, the meaning of a sentence can be broken down into propositions, which are small units of knowledge. For example: the sentence 'Susan gave a white cat to Maria, who is the president of the club' is broken down into three propositions:
1) Susan gave a cat to Maria.
2) The cat was white.
3) Maria is the president of the club.

Anderson also proposes that every concept in a proposition can be represented by a network. He argues that the spread of activation in working memory has a limited capacity as there are 10 nodes represented in working memory only. Thus, if there is too much activity going on, each link receives relatively little activation.




No comments: