Experiment 2: Simple Emotion Modelling

Ditto the donkey is beginning to learn the basics of the English language. He currently knows about 1,000 words, and his vocabulary is gradually growing as people interact with him. However, he can't reply to you, as he hasn't yet started to learn how to speak.

This experiment has been set up as a simple example of supervised machine learning and affective computing. Using a training set of examples which reflect nice, nasty or neutral sentiments, we're training Ditto to distinguish between them. At this early stage he's seen only a few examples, so we'd like you to help us by providing some more.

When Ditto receives a message from you, he evaluates it for niceness or nastiness, then responds emotionally. He has a scale that runs from -100 (desperately unhappy) to +100 (ecstatically happy) and you can see this while you're talking to him. The more you insult him, the more miserable he'll get; the more you compliment him, the more he'll cheer up. However, it may take a few messages for Ditto's emotional state to change much. Try being nice to him for four or five messages, then nasty for a similar number of messages. You should see his mood change dramatically!

Ditto isn't perfect, and he may make mistakes. We hope he'll improve as he becomes more experienced.

We teach Ditto away from the website. Therefore he won't learn instantly from things you say.

Ditto was last trained on 25 July 2007, using 7060 examples.

We're logging your messages and we'll use them as a source of training examples. However, you can't be identified as an individual, so feel free to say whatever you like. Please interact with Ditto and help him to learn!

Experiment 2: Simple Emotion Modelling

Links