The Enigmatic World of Wikispeedia
Our minds are complex instruments, constantly composing thoughts, yet we know so little about the notes behind the music. Usually discreet, our brains occasionally let the curtain slip, revealing their inner workings in unexpected settings. Enter Wikispeedia: a cognitive revelation hidden in a simple game. Players dart from one Wikipedia article to the next, aiming for a target. As they navigate, they leave behind a trail of thoughts. What secrets do these trails unravel about the voyagers? Take a trip with us as we explore these mental pathways.
Game Characteristics - Cracking the Wikispeedia Enigma
To crack the code of player behavior, we first need to understand the game’s properties. In Wikispeedia players journey through a maze of Wikipedia articles 4604 articles spanning 129 semantical categories from Geography, History, Technology, you name it !
In an attempt to connect the dots between a source and a target article, players have weaved 405,835 paths :
- 256,585 successful attempts paths where the target was reached
- 149,250 unsuccessful attempts
Since we’re trying to leverage the player’s behaviour to map their cognitive patterns we will see articles in paths from the perspective of their specific categories and find the connections between categories that appear in that setting.
However we need to take a stop and analyze what makes a difference between successful and unsuccessful players. Do successful players simply “know better” or is there something deeper in the works ?
What’s the game plan coach ?
The first layer of a player’s knowledge is their strategy. Imagine trying to go from point A to B in an unfamiliar town (no GPS allowed of course), how would you proceed?
You’d likely try to find your way to a central location like the town square, and proceed from there. Something very similar is observed in the basic strategy that players, whether they finish the path or not.
To uncover this strategy we need a measure of the “popularity” of an article. However, we should take into account the unbalanced nature of our links, the in-link out-link distribution is different between the two :
Simply basing our popularity approach on the degree isn’t sufficient, simple metrics like counting degrees may not capture the nuances of the network’s structure. This is where the PageRank algorithm enters the fray, providing us with a robust measure of an article’s stature within the game.
Path Length: 4
As we suposed, players take a hike, find a good “peak” and try to go down as fast as possible to the target, when they fail, they repeat the procedure again, until they succeed or get too tired of hiking !
Since this strategy is common to both successful and unsuccessful players, we need to dig a bit deeper to see if there is a true difference between them.
Luckily now that we have the clear distinction between the two phases of the game :
- Up Path : looking for the highest first peak
- Down Path : going down from the peak to find the target
We have isolated where player cognition (and connections between categories) come into play : the risky Down path, here players show off their knowledge and navigationnal skills, we define a proximity metric between categories based on how often they appear in succession in a path, meaning they are not only tightly connected in the structure of the game (since the link between the two is present) but the player knowingly chose to take them therefore justifying their use as a connection index, we can observe the following connections :
- Heat map of the finished Down Paths:
- Heat map of the unfinished Down Paths:
Aha, the patterns are different ! But wait these plots actually prove that unsuccessful players explore more connections that succesful ones, hence play success is independant of strategy and knowledge. it is maily related to the game instance.
Now we know that all paths contain breadcrumbs to player cognitive connections, all that is left now is to follow them :
Extracting the Juice
In order to get a better intuition on the proximity of the categories, we construct a graph based on the adjacency matrix of the heatmap. We see obvious patterns that we see in everyday life: When I say history, you probably think geography! When I say art, you probably think of a certain artist so the people category.
Now let’s delve into the learning behaviour of the players. In first game plays, players tend to explore various strategies without a clear and efficient pattern.
The idea here is to visualize the evolution of the semantic links through a metric we defined depending on the weights of the links between categories (found in the heatmaps). This metric is then used to extract the distances separating the categories to create a scatter plot that translates the semantic closeness of the categories.
We take care of removing the outliers keeping only the players that played between 30 and 250 games.
We compare first plays Before training against last plays After training to gain insights on the learning process of player, we present here the evolution of 4 of these categories :
The pre-training plots show a central node surrounded by many nearby nodes, indicating a stage of exploration. The proximity of these nodes to the central one suggests an initial understanding of the game but without optimizations (tricks and shortcuts).
In contrast, post-training plots illustrate two significant changes :
- There are fewer nodes near the center target, but they are much closer! This shows that players are now focusing on more effective strategies that bring them closer to their goals.
- There is an increased distance of the other nodes from the center reflects a discard of less useful paths.
Essentialy, when we look at the progress of experienced players, we can clearly see how they improve from their initial tries to their more advanced techniques later on. This shows us how their their cognitive map of the game sharpens !
The Odyssey of an Exceptional Player
But there’s an outlier in our adventure: a player who’s navigated through about 5000 games. What’s the story behind this marathon? This player is like a rare gem in our exploration, offering a unique perspective into the intricate workings of a seasoned Wikispeedia navigator. In our analysis, we use p-values and t-test values to examine a key hypothesis: Does the time taken to find the final destination in a given category vary significantly after the player has played many games? A smaller p-value (at the 0.1 threshold) suggests a significant change in time spent, on the other hand the t-value indicates the nature of this change:
- Negative t-values mean the player is becoming slower
- Positive t-values suggest the player is learning faster, mastering the category quicker
In IT and Geography, the significant p-values and negative t-values indicate they are taking more time in these categories. This slowing down might be due to a more methodical approach, perhaps seeking a decision to focus on accuracy in complex areas
However, for Design & Tech, Religion, and Language & Litterature although
initially slower, have shown over time remarkable improvement.
It’s like unlocking a secret power-up, making leaps in understanding!
So, our intrepid Wikispeedia player, initially more lost than a tourist without a map in Design & Tech, Religion, and Language & Literature, has turned into a seasoned navigator. It’s like they’ve cracked the code, moving from ‘Where am I?’ to ‘I’ve got this!’ in record time. They’ve transformed their journey from a meandering stroll through Wikipedia’s corridors into a high-speed chase for knowledge, outpacing their past self with the finesse of a trivia maestro.
Conclusion :
In conclusion, our exploration of Wikispeedia unveiled fascinating insights into the cognitive processes of players expressed as the connection between the categories they choose. While they shared a common initial strategy, the “Down Path” phase highlighted their ability to form meaningful cognitive connections between related categories. Their learning behavior was evident in veteran players, who mastered unfamiliar subjects over time. Essentially, our journey showed the dynamic nature of human cognition and the richness of knowledge acquisition within the Wikispeedia game.