Enoch Fwa Undergraduate Dissertation 2015/16

Scikic: The Artificial Psychic (in potential collaboration with The Good Thinking Society)

Supervised by N.Lawrence

Abstract

The project involves building an online inference tool called Scikic . The online Scikic is a website, which takes a set of question/answer items and then queries a series of local and remote databases to generate conditional probability distributions over various features. Currently, age, gender, sex, location, and certain insights on their environments. The API is highly modular, and some modules don't use the probabilistic framework rather provide some useful information. The conditional probabilities are combined using a Bayesian network, using the pyMC module. Each module can provide pyMC 'features' which create functions to output the relevant probability distributions.

  The main initial task is to build the artificial psychic’s information engine based on probabilistic inference. This would be a general framework for selecting, creating questions, recording answers, incorporating new datasets calculating conditional probabilities and performing inference on these conditional probabilities. Probabilistic inference is a great approach to reasoning uncertainty and empirical learning.

  The software for the project is written in python according to the principles of open data science (Lawrence, 2014) in potential collaboration with ‘The Good Thinking Society.