An algorithm that works similarly to Grammarly, except instead of how well written your text is, it shows you how original your expressions are.
It's more difficult to keep being boring when the originality score is straight in your face each time. Coming up with something novel requires putting some thought into it, which equals additional time and energy, this is the reason people rarely try being original. Displaying originality level is a passive-aggressive way to encourage this.
People will rarely tell you that you're being boring or/and repeating yourself, because of the fear of hurting your feelings. An algorithm, on the contrary, is a different story.
Might help enliven stagnant parts of relationships, especially those with family members, close relatives, partners, etc. where the same topics are frequently touched, but the discussion then moves nowhere and both sides tend to reply with similar phrases each time. The latter was what inspired the idea.
It could function as an extension for chat apps, emails, social media pages, etc. The reference point for originality is your chat history with a particular person. If you only start chatting with someone, the algorithm will compare your first sentences with those you used when starting to chat with your other contacts or/and it could take general human communication as a reference point in which phrases like "hi, how are you?" would be scored the lowest originality level since they're extremely common.
This is not to say that you shouldn't use common phrases or small talks in your speech, after all, you have to start with something, you could simply ignore the originality score at first, but if your score keeps being low as the convo progresses, you might want to consider diversifying your speech.
The algorithm would start displaying the level of originality as you type the phrase so that you could consider if you really want to send it before the "damage" has been done. It could also suggest alternative phrases/expressions to make the language more rich and original.
But the main focus should be not on how you arrange the words and what synonyms you use, but on the meaning of your expressions, if you change some words for the sentence to sound differently, this might help a little, but the essence is the same, you might want to talk about something entirely different, answer in a different manner than you usually do, only then it can be considered more original than your usual interactions.
How would the algorithm measure the originality level?
I'm not a programmer, so I can't suggest the mechanisms behind it, but it doesn't seem like a hard task to do for an algorithm. Even counting the repeating (as compared to past conversations) words and phrases in sentences and basing the score on the level of repetition, would give a productive result.
I didn't find sources on the web suggesting how to measure the originality of speech, but I did find those suggesting methods to measure the novelty of products, ideas, and scientific discoveries, that can all be helpful in attempting to design the proposed algorithm.