Connections and Sparks: Network Analysis Meets Speed Dating
Decoding Speed Dating
Have you ever wondered how speed dating works behind the scenes? It’s not all random — there are patterns hidden beneath the surface. And I’m here to reveal them!
Using network analysis, I’ve converted the whirlwind of speed dating into a visual story of connections. This isn’t just a who-likes-who map. It’s a deep dive into the dynamics of attraction, the role of gender, and the subtle sway of attractiveness.
Join me as we explore the world of speed dating, not as a chaotic flurry of faces and names, but as an intricate network filled with insights. Ready to see the unseen? Let’s get started!
Date-a Selection
For our exploration, we used a unique dataset collected by Columbia Business School professors, from a series of speed dating events held between 2002 and 2004. Each participant had a quick four-minute date with every other participant of the opposite sex, rating them on six attributes, including Attractiveness and Sincerity.
The dataset is rich with a range of data points, from demographics to dating habits, but we focused on key elements that make our network analysis meaningful: information about the participants, who they dated, whether they matched, their ratings from their dates, and if they were liked in return.
We analyzed the data from Wave 17, which took place on February 25, 2004. This wave was chosen due to its large number of participants and the completeness of the rating data.
Painting Love with Graphs
To unravel how connections spark in speed dating, we’ll dive into the world of graphs, a simple way to illustrate who’s connected to whom.
Our study will feature two kinds of graphs — ‘undirected’ and ‘directed’. Both are built using a concept called ‘bipartite graphs’. A bipartite graph separates nodes into two sets, in our case, males and females. Connections only happen between these sets, mimicking the speed dating setup.
In an undirected graph, a line (or an ‘edge’) between two people means they both liked each other, forming a mutual match.
But what if only one person felt the chemistry? That’s where directed graphs come in. Here, an arrow emerges from one person pointing towards another if they liked their date. This can be made more complex by adding an extra dimension. The attractiveness rating influences the arrow’s transparency. The more attractive someone found their date, the clearer their arrow, signaling their strong interest.
Our First Dance: The Undirected Bipartite Graph
In our graph, the orange represent females, while the green represent males. As our first exploration, we’ve created an undirected bipartite graph, a perfect snapshot of mutual matches.
A quick glance shows us that Kendra, Robert, Jorge, Ricardo, and Leroy stood alone, no connections drawn to them. This might suggest no one was interested, but remember, we’re only seeing part of the picture — mutual matches. What about those one-sided attractions, those hopeful individuals who were drawn to them? To unravel those hidden stories, we’ll need to dive deeper.
Unveiling the Number Game
At our speed dating event, we began with 10 females and 14 males. Women on average made 2.4 matches, and men made slightly fewer, around 1.7 matches. Connections varied wildly, with Sarah, making a remarkable 6 connections, while a few ladies made none to one match. The male participants followed a similar trend. David, a sociable gentleman, made 5 connections, while several men had no matches.
If we consider the degree distribution — the frequency of different numbers of matches — we notice distinct patterns. Female degree distribution looked like a normal distribution, although slightly leaning to the right, indicating that most women had an average number of matches. However, for males, the distribution was skewed to the right, meaning a significant portion had fewer than average matches, with some men forming several connections.
Our graph is a bipartite one, indicating that matches were only between different genders. To understand inter-gender relationships, we created ‘projections’ — separate graphs for each gender where edges mean they matched with the same person from the opposite gender.
When we computed the Clustering Coefficient in each projection — a measure showing how much one’s matches are similar with other participants in their set — females had a higher coefficient (0.67) than males (0.53). This suggests that if two women had a mutual match, it’s likely that another woman also shared a match with either of them. This is interesting because it could indicate that the qualities that make a man attractive to one woman are likely to make him attractive to other women as well, and that men’s preferences appear to be more diverse or heterogeneous, suggesting that the qualities they find attractive in women are more varied.
Lastly, the Connected Components, revealing our speed dating event resulted in 7 distinct groups or networks of inter-matching individuals. It’s a reminder that while participants may find a match within their group, they had no matches with individuals in other groups.
Remember, these insights are just about mutual matches and don’t capture the full story. But they offer an intriguing peek into the attraction patterns during our speed dating event.
Looking into Directed Graphs
Let’s dive into the intriguing world of directed graphs, which gives us a more detailed understanding of the patterns of ‘likes’. Here, we consider both ‘in-degree’ (the likes received) and ‘out-degree’ (the likes given), to provide a more nuanced picture of our participants’ preferences.
In our speed dating setup, we had 10 females and 14 males. An interesting pattern emerged: the females, on average, received likes from 7.4 men (their average in-degree), yet expressed interest to only about 4.7 men (their average out-degree). Conversely, males received likes from about 3.4 women, but gave likes to around 5.3 women. This suggests that females were more selective, yet more desired.
However, the picture changes when we introduce ‘weighted degrees’, which take into account the intensity or ‘weight’ of these likes. Despite expressing likes to fewer men, women gave these likes a higher intensity on average (7.1) compared to men’s average (6.07). This indicates that while women were more selective, when they did express interest, it was often more intense. Similarly, the average intensity of likes received by women (6.8) was also higher than that received by men (6.5), suggesting that the attraction towards women was generally stronger.
These observations from our directed graph reveal the fascinating dynamics of the speed dating event, with the level and intensity of mutual attraction varying across genders.
Deeper into the Dating Game
Diving deeper into the degree measures, we found some compelling trends. The four male degree histograms (in-degree, out-degree, average weighted in-degree, and average weighted out-degree) were skewed to the left. This indicates that, while most men received and expressed more ‘likes,’ there was a smaller group of men who received and expressed far fewer. When we look at the undirected graph, it’s evident that a larger number of men had fewer matches. This contrast could be due to the fact that the men who were highly desired by women (contributing to the high in-degree and left skew in the directed graph) did not necessarily express interest in a lot of women (resulting in fewer matches and a right skew in the undirected graph). It suggests that the men who received many ‘likes’ from women were selective, reciprocating only a few of these ‘likes’.
For women, the in-degree histogram was roughly a flat normal, indicating an even distribution of ‘likes’ received, suggesting that women on average were desired almost equally. The out-degree histogram was right skewed, indicating that most women expressed interest in a smaller number of men, but a select group expressed a high number of ‘likes’. This agrees with the earlier statement that women were more selective. The left skew in the average weighted in-degree and out-degree histograms for women illustrates that a number of women both received and gave high attraction scores, showing that there were a lot who were particularly desired and who also found many men highly attractive. But, if a lot actually gave high ratings (average weighted out-degree), but a lot gave out few likes (average out degree), this agrees to the observation that while women in general give fewer likes, when they like someone, they like them bad.
Degree centrality, which showcases a participant’s immediate influence, also presented distinctive insights across genders. These distributions highlight the varying degrees of influence within the dating network.
Across the Table
The data from the speed dating event provides us with a colorful weave of interpersonal connections, presenting a fascinating illustration of human behavior. These connections are varied, with some individuals forming mutual likes with several others, while some had difficulty finding a single match. This is a reminder of the inherently personal nature of dating, a process driven by individual preferences, which can often diverge significantly.
One notable aspect of our analysis is the clustering coefficient. The clustering coefficient in our study measures the likelihood that two people from the same set, both of whom have expressed a ‘like’ for the same person from the other set, are also similar with another person in their own set. This means that if two women, for instance, both ‘liked’ the same man, the clustering coefficient will indicate the likelihood that these two women also share a ‘like’ with another ‘woman’ for the same men in their dating pool. This provides an indirect measure of shared tastes or compatibility. The higher clustering coefficients indicate that there’s a significant level of shared attraction within the group, reinforcing the idea that certain individuals or attributes may be widely appealing.
Gender, however, emerges as a key determinant of dating strategies. Men, as perceived in our system, appear to express ‘like’ for more women than the other way around. This suggests a broad, non-selective approach, perhaps based on the principle of maximizing potential matches. Women, on the other hand, appear to adopt a more selective strategy, expressing ‘like’ for fewer men. This selectivity, though, doesn’t translate into fewer likes received — quite the opposite. Women receive more likes overall, reflecting a gender dynamic where men appear more proactive in expressing interest.
What’s also interesting is that when women express their ‘like’, they do so with more intensity, as evidenced by the higher attraction scores given out by women. This might suggest that women, while being more selective in choosing whom to ‘like’, express a stronger attraction towards the men they do like.
The nuances don’t stop at the gender level. Individual strategies and outcomes also showcase a wide variety. Some men, a small group, stand out for receiving and expressing more ‘likes’ than their peers, suggesting high engagement and possibly, higher desirability. This group presents a contrast to the majority of men, underscoring the heterogeneity within gender groups.
The data also reveal intriguing individual stories. Kendra, Jorge, and Ricardo stand out as they received high in-degree centrality scores but had no mutual matches, hinting at a high degree of selectivity. They received interest from others but didn’t reciprocate, indicating personal preferences at play.
Leroy paints a different picture. While this individual expressed interest in others, the feelings were not reciprocated. Conversely, the individuals who liked Leroy were not the ones he was interested in. This dynamic underscores the complexity of mutual attraction.
The most miserable case is perhaps Robert, who didn’t receive any ‘likes’. This could be due to various reasons: Perhaps they didn’t appeal to this particular group, or maybe their perfect match wasn’t present at this event. Let’s hope he found his match in another wave of speed dates.
Taken together, these insights paint a detailed picture of the complex interplay of factors at work in a speed dating event. Personal preferences, gender-based strategies, and the subjective nature of attraction all contribute to varied outcomes. These results offer us a peek into the intricate world of dating, where individual choices, strategy, and luck intermingle to weave unique narratives of connection and missed opportunities.
Networked Romance
We used two kinds of ‘networks’ or ‘graphs’ in our study: undirected and directed graphs. Imagine undirected graphs like a two-way street where ‘likes’ go both ways. They showed us who liked each other in the speed dating event. This was a simple and fun way to see who had mutual feelings. But to get a fuller picture, we needed more.
That’s where directed graphs come in. These graphs are like one-way streets. They show who likes who, even if the other person doesn’t feel the same way. It’s a little sad, but it also lets us see the full story. These graphs told us more about what happened in the speed dating event — who was popular, who was selective, and even who was unlucky this time.
Using network science, we made speed dating feel less like a game of luck and more like a cool science experiment. It was like watching a social network unfold in fast-forward!
As we wrap up our exploration of love through the lens of network science, remember that speed dating, while it can be analyzed and studied, is not an exact science. It’s a whirlwind of human emotions and surprises. So dive in, enjoy the adventure, and who knows? Your perfect match might just be a ‘like’ away.
