Week 4 - Surprise, Evaluation, ReflectionΒΆ

  1. RETURN The creative artifacts are often said to have three features: high value, novelty and they are surprising. In the last weeks’ exercises we have built means to compute the value and the novelty, but have not directly measured surprisingness.

    Give a short explanation of what a surprising artifact from an agent’s point of view would be, and a few examples of how the evaluation of surprisingness could be implemented in a simple way. Write your answers briefly (10-15 sentences).

    Additional notes:
    • Some of the functionality implemented in the course might be already suitable to measure the surprisingness. You are free to use them as examples. Also, keep in mind that value, novelty and surprisingness are not actually orthogonal measurements of the artifact, i.e. the measurements that measure value may also measure surprisingness to some extent.
    • Wikipedia’s definition for surprise is as good as any: surprise (emotion), a brief emotional state experienced as the result of an unexpected significant event. Here, the event would naturally be an observation of an artifact.

  2. RETURN Pick one from your examples of the implementable suprisingness measures, and implement it to your agents. What are the main differences between it and the value and novelty measures? Is there any overlap between them?

  3. RETURN (3pts) Consider your current agent implementation (with all the functionalities you have implemented in the exercises during the course). Explain briefly your current agent design and the functionality of the multi-agent system as a whole and answer the following questions:

    Ritchie (1pt) (see the article by Anna Jordanous for short introduction):
    • What is the inspiring set (if we think as an inspiring set the corpus from which the Markov chain was learned) for a single agent and the multi-agent system as a whole?
    • Is the agent (or society) able to reproduce all instances in the inspiration set? If not, give an example.
    • Is it able to produce instances outside the inspiration set? If so, give an example.
    • Is it able to produce valued instances outside the inspiration set? If so, give an example.
    Ventura (1pt):
    • On which “level” in Ventura’s article does your single agent implementation land? Why?
    • Consider now the multi-agent system as a whole where the output of the system is the set of vote winning artifacts (one per iteration). On which “level” does the system as a whole land? Why?
    Wiggins (1pt):
    • Explain your single agent implementation with Wiggins model. What are \(\mathcal{U}\), \(\mathcal{T}\), \(\mathcal{R}\) and \(\mathcal{E}\)?
    • Explain your whole multi-agent system implementation with Wiggins model. What are \(\mathcal{U}\), \(\mathcal{T}\), \(\mathcal{R}\) and \(\mathcal{E}\)?