Hello fellow founders & co founders. I have designed an mobile application that will feature a algorithm that will take the users data from a 36 questionnaire and tag each profile with a certain color that will correlate to a particular psychological profile. Each of the users will be asked to answer particular questions from Myers brigs test, Attachment theory test and 5 love languages test and the app will generate a color. The app will suggest matches that are 80% or better.
What would you recommend is the best way to go about this?
5 answers:
Hi Alexa,
I also believe that the world needs something newer than apps like Tinder and have studied the market extensively.
Personally I do not believe there is any future in AI dating apps. The reason is that not only is it inefficient to calculate potential love stories with technology, but also a main issue is that if by miracle you have developed the love making AI algorithm, even if you become the world's biggest dating app like tinder, this would still leave users with very few matches. You are constrained by the fact that users must be close in location and be active on the app. Also consider that out of 30 matches on Tinder, if 2 lead to something you are already lucky. And to get 30 matches on Tinder the average person will have to swipe an extensive amount of people. Therefore, narrowing down matches so much with AI means your users will get very little movement.
And to even get there you would have to gain 10 of millions of users in an already crowded market. Your "screening process" exists in similar apps which have failed due to the reasons I stated above. Do not get mislead by the amount of downloads of an app, it does not correspond to the active user base.
To improve online dating I would advise going in the opposite direction by focusing on finding a way to capture the widest worldwide user base possible. People are good at deciding who they want to go on a date with. Current apps lack enough users for people to go on as many dates that they would like.
To conclude, focus on scalability potential for a dating app, not match quality. That can be introduced when you have 500 million users, but you can't do it the other way round.
Hi Alexa, The approach you are taking is quite innovative however I believe it will lead you to what most dating apps currently cater too i.e profiling based on set information provided by the user. This is how traditional matchmakers work as well. A different approach would be, analyzing how users are spending their attention and drawing similarities to others who also have similar patterns and try to provide recommendation. The point here being humans are not static beings, and long questionnaires need not necessarily be the correct responses. Instead if you are able to tap in other subconscious behavior patterns, that might be a better predictor of people having similar wavelengths.
Hey Alexa,
The best way I see something like NLP being incorporated into a dating app would be within the chat feature. It could help generate questions regarding common interests, or come up with ideas on where to get lunch. This "wingman" feature could help keep conversations flowing as people attempt to create a connection.
I don't quite see where the AI fits into this. If I understood you correctly, what you are proposing is essentially a rules based algorithm. Rules dictate the mapping between answers and colours and presumably rules dictate which "colours" go well together.
AI and machine learning is the ability to generalise based on examples. In your case, you would identify N couples who's relationships are going well and N couples that didn't work out so well. You would feed everything you know about the individuals into "the machine" and it will build a model. You would then pass the profile data for two new individuals into the model and it will make a prediction as to whether their relationship will work out.
I will discuss your questions from two aspects: (1) Is your method of collecting data reliable? (2) How does mainstream AI do it?
(1) When you deliberately use a questionnaire or method to collect users' ideas, you will be questioned by statistical experts. For example, the reliability and validity of questionnaire design. Even the information collected through a rigorously designed questionnaire survey can only be used as a reference.
For instance, when we want to ask questions to investigate "how many people have you been in love with", "the number of times you have sex per week", or "how much time do you spend on sex". As we know, it’s a fact that people usually lie on such questions.
So when researchers need the data, they will cooperate with medical institutions. They will use the medical institution's medical records (anonymous statistics).
You mentioned that “Each of the users will be asked to answer particular questions”. This method will be questioned by experts. For example, how do you prove that all people being tested will not lie?
(2) There are many routes to realize AI, including Machine learning, Data mining, Traditional AI(game tree), etc.
For “Machine learning” or “Data mining”, big data is necessary. Big data may be a record of people's various activities. For example, a daily conversation between friends chatting through twitter, shopping records (books, food, merchandise, sex toys, etc.), and travel records, etc. After the AI has analyzed the big data of tens of millions of people, when you use this app, the system may recommend you: “Among the crowds, there is a man/waman worth your time to be friends with. You may regret it if you miss him/her.” Maybe this is more convincing.
Finally, the age of quantum technology is about to come. Quantum technology to achieve AI will be another route.
The picture below is Germany's first quantum computer.
