WHAT AI HAS TO SAY ABOUT:
GAME OF THRONES –
THE LONG NIGHT


Analysed by Soroosh Mashal & Pascal Hecker

Some Scars Never Fade

It was 2:00 AM, and my phone rang. “Get down in 15 minutes. I will pick you up”, my friend said. “Why? What? What’s going on?”, I replied. “New Game of Thrones episode. We have to watch. NOW!”, he answered in broken sentences.

It was 3:00 AM, I was staring at the black screen and the credits were passing before my eyes. There was not even any music to mend the wounds from the “red wedding”. It took us some hours to comprehend it thoroughly, however, the scars have not faded, yet.

Playing with Our Toys

“The long night” was released. Everyone knew it is going to be big. Normally, we watch new episodes with other colleagues together. However, this time was a bit different. We decided to use some of our tools on our colleagues because we were expecting some deep reactions.

“What if we use the artificial intelligence to assess the level of excitement in the reactions?”, I said.

Well, I think you can imagine what happened next. We placed microphones in the conference room and recorded the audience throughout the screening. Then we passed the acquired data to ExamAIner which is a web application using the emotion recognition audEERING technology in the backend.

Analysis and Results

Using this method, we could measure all emotion levels for each of the reactions. Thus, we came up with a graph for the level of expressed excitement throughout the episode.

To understand this analysis, we must first see how we react to such a media. Normally, you tend to watch a movie in silence. We will start expressing ourselves, when we already pass an excitement threshold. The way and the amount of expression is dependent on factors like environment (e.g. how comfortable you are), personality (e.g. extrovert or introvert) and many more. These factors can change the rate, amount, and threshold of our expression.

audEERING analysed Game of Thrones

In this experiment, we detected 184 expressions during the episode. The correlation between important events and the amount and intensity of expression can be seen clearly in the graph. To keep this post spoiler-free, we will not go into the details of each scene, however, three major peaks have been identified in a way that people would understand what those peaks refer to.

Now, I took this data, and send it to my friend who picked me up to watch red wedding some years ago. This time I could share exactly how exciting this episode was without even spoiling a single scene.

I know it might sound really nerdy, but we love to play with our toys. Besides that, imagine the applications of such tools for fields like market research. Using such tools, you can tap into the most natural and basic responses and reactions of your audience.