In incremental updates of the goal likelihood (line 351 to line 353 in file Gamydala.js), likelihood is calculated as:

	newLikelihood = oldLikelihood + likelihood*congruence; //line 352
	newLikelihood=Math.max(Math.min(newLikelihood,1), -1); //line 353

Like me a miss calculation of the goal likelihood in a range out of [0, 1] (line 353). A fix for this issue can be:
newLikelihood=Math.max(Math.min(newLikelihood,1), 0); //line 353

Right?

Thanks

Consider the following code:

let emotionEngine = new TUDelft.Gamygdala();
let emotionAgent = emotionEngine.createAgent('player');
emotionEngine.createGoalForAgent('player', 'win', 1.0);

emotionEngine.appraiseBelief(0.5, null, ['win'], [1]);
emotionEngine.printAllEmotions();

emotionEngine.appraiseBelief(1, null, ['win'], [-1]);
emotionEngine.printAllEmotions();

Output:

player feels hope:0.25, 
player feels hope:0.25, distress:0.75,

We would expect an agent to feel hope after the first appraisal, as they do here, but not after the second appraisal once we've shattered their dream of winning. However, the "hope" emotion is still hanging out in the agent's array of emotions.

I'm not sure if this is a bug, a limitation of the system, or intended behavior. It doesn't seem to make much intuitive sense to me, but I admit I might be missing something. Is the current emotional state supposed to be generated anew for each appraisal? Or, are the old emotions supposed to hang around, as happens now, even when they seem incongruous?

I believe the agent should also be feeling disappointment: 0.75 after the second appraisal, but that issue is mentioned here: #9

Thanks for a great example of Gamygdala implementation.

Trying this out in a translated version for SWIFT to run on Mac (done myself based on your JS code) I have a few questions

1. It seems the model can have the same goal associated with different NPCs (share goal) - however, is it correct that utility cannot vary in the implementation ? Let’s say two NPCs has the same goal but that one has disutility (negative utility) from achieving the goal - and the other has utility from achieving the goal. It seems that both NPCs would refer to goal identified by goal-name uniquely.

2. My other question is about calculating internalEmotionalState where the condition to become "disappointed” is Likelihood of achieving goal equals zero and secondly that the change in delta likelihood is strictly above 0.5. This implies that LH must have been negative before the belief-update - but that does not seem to be possible since the normal update rule for new Likelihood = (Congruence x Likelihood + 1) / 2 - thus a number between zero or 1. Thus, it seems that “disappointed” emotion can never be reached? Am I missing something here? Should deltaLikelihood be evaluated as below -0.5 - or maybe just as absolute (so only extreme values) ?

I fully appreciate that this is an older project so I don’t expect an answer.

Thanks for a inspirational article and very thorough implementation,