Am I Happy ? works using word association. Essentially, it aims to understand how you relate to certain words using other carefully selected words.
Essentially, the word that you select can be one of three things : associative, descriptive or emotive.
Associative: In this case, you select a word that is associated with the concept. So, let's say the word is star. You might choose the word night. It isn't a word that exactly represents the concept but is associated with the concept directly.
Descriptive: This would mean that the word you choose describes the concept. It can be broken into two sub-sections - abstract and representative. In the case of representative, then it would be a direct statement of what you are considering. For example, if the word was again 'star', then a representative description could be 'sun' or perhaps (incorrectly) 'planet'. It attempts to describes the concept (or object in this case). An abstract descriptions are more complex but let's say, for example, you are considering the word 'star' again, you might use the word 'twinkle' or 'shine'. You are describing the concept but not in it's literal sense. It's an abstract description of what a star is and how it appears.
Emotive: An emotive word would be one where you state how you feel about that concept. So, for example, for the word 'star', you might choose the word 'wish' or 'distant'. In some cases this can be similar to associative words but the difference is that it covers a feeling that you might have. The concept of a 'star' might make raise thoughts of distance. Emotive responses need to be related to the word in question to be of any use. For example, an emotive response to star could be 'happiness' itself. But clearly, such a response could be from any word. Emotive responses can, of course, be positive and negative. Being 'afraid' in response to a word concept is clearly an emotive response but probably in a negative way. That doesn't mean that you are not happy - quite the opposite. Being in touch with emotion is often considered being a key to happiness.
This is where it gets tricky. We used an advanced Neural Network to understand the words inputs relating to the words in question and the responses to all fifteen words. The words themselves are graded using different techniques and the frequency for which they appear and also the the frequency across words. For example, a similar response across all word concepts might be a generic emotive response and therefore carry little weight to the overall score. A change from a trend can often be of more validity.
Different responses to different images are representative of how happy you are. We break down the different responses but essentially it can be summarised that emotive responses reflect a person more in-touch with their feelings than someone who simply uses descriptive responses. That doesn't mean that you are happy - it could equally mean that you are sad. So the responses themselves are analysed to understand which way you are likely to go.
It's worth mentioning that the system is trained using a large set of data to create the Neural Network. In reality, we don't even know why the system moves some people towards happiness and others towards sadness. All we know is that the results are usually interesting !