Researchers have developed an online dating system that not only matches you with partners you’ll find attractive, but who are also likely to find you attractive too.
The researchers at the University of Iowa have addressed an underlying problem of online dating sites. There’s no doubt that such sites are ever increasing in popularity, and have good algorithms taking into account the reported likes, interests and hobbies of the person looking for a partner to come up with a potential match. What’s less well catered for is the trickier aspect of the reciprocal interest – you may think person x looks nice, but will they find you equally attractive?
Most online dating software works out whether you’re likely to find someone attractive by analyzing the people you’ve chosen to contact to see what interests and tastes they have. This is then used as the basis for working out new recommendations on the basis that ‘people who liked this person also liked these people’. The researchers say (in a paper published by the IEEE), that “the model considers a user’s “taste” in picking others and “attractiveness” in being picked by others.”
The problem here is that if you are Average Joe and try asking out Supermodels Ann, Barbara and Cheryl, you’re unlikely to get a reply. Well, not a printable one, anyway. So coming up with yet another supermodel for you to sob over isn’t a lot of help.
Instead, the researchers add a note of reality by analyzing the replies you get, and use this to work out how attractive you are. This is a scary thought for many of us, and one we may well not want an honest answer to. The results are used to recommend people who might actually reply if you get in contact with them.
The theory behind this has been tested out using data from which names and identifying data has been removed, drawn from a dating website with 47,000 users over 196 days. The data from the first 100 days is used as a training set to calculate the attractiveness and tastes of each user. This was then used to create recommendations from the remainder of the data set, with other selection criteria for comparison being just what the users chose, or by matching other variables such as each person’s likes and dislikes.
In each case, the researchers first looked at the number of potential dates recommended by the three methods, and analyzed how many potential dates the users contacted to estimate how good the methods are at estimating whether the user will find potential dates attractive. Next, the researchers looked at how many of the contacts actually replied as the measure of how suitable the date was. The researchers say the results show that “the new model has good performance in recommending both initial and reciprocal contacts,” commenting that “if a user approaches a partner recommended by [our software], he/she will have a better chance of getting response.”
Fortunately for the attractively challenged, the research is still just that – research.
However, given the fact the online dating market is worth around $3 billion a year, chances are someone is going to make use of this.
We have been warned.
User Recommendation In Reciprocal And Bipartite Social Networks–A Case Study Of Online Dating by Kang Zhao, Xi Wang, Mo Yu, and Bo Gao
There is still time to come up with an idea that could win you big bucks in the Intel RealSense App. If you want inspiration there is now plenty on offer and you can also reserve a RealSense developer [ ... ]