There’s a lot of talk about the firehose of data. This constant stream of information that we seen pouring out of every device that can produce it. The Twitter firehose is an obvious one as there’s so much data coming out of it.
Marketers love it because they can measure sentiment.
Sentiment is a very strong word and one that I feel has been slightly misplaced. The basic dictionary definition of “an attitude toward something” slightly quashes the more important root of the word from the medieval Latin “to feel”. And in most respects the feeling with data is lost.
Data sentiment is usually measured in the most rotten awful way from my previous experiences. The common method of taking a positive and negative word lists and running every sentence past it to create a basic score is just that, basic. To see how easy it is you may want to read a previous blog post I did.
The sentiment of data and the sentiment of the person can be two different things. Our online lives are sometimes a shallow mask of what we want people to see. Not always how we really are. Below are public tweets with the words “I feel like” contained in them.
[33] “Getting introduced as “My Girl” should be a positive thing, but I don’t like the label, and now I feel committed #Confused”
[34] “RT @ThereGoesThtMan: Feel like I’m waiting for something that isn’t gonna happen”
[35] “Wow. I feel like I’m in middle school again watching this show. #lagunabeach”
[36] “@JLSOfficial I feel like I’m talking to the zombies when I tweet you. Haha”
[37] “When somebody lies to me once I feel like everytime they speak they’re lying…”
[38] “RT @_jasminekovacs: They told me you was trouble but I never knew I’d feel like this”
[39] “I feel like I’m going to die”
[40] “I feel like I do have a right to be mad tho . NIGGA DIDN’T ANSWER ANY OF MY CALLS.?! Tf”
[41] “I don’t feel like goin to work today”
[42] “I don’t feel like getting ready today.”
[43] “RT @FemalePains: Hurts to even breathe right now. Feel like I’m about to throw up. #femalepains”
There’s a whole mixture of mood, statement and feeling. It’s difficult to really get to the bottom of fast.
The other issue is that you’re normally dealing with one source of data whether that be a tweet, a Facebook update or what have you. Mood on the other hand and it’s correlation with how we feel is a much more challenging concept.
Do certain places made you more upset than others? Does eating certain foods trigger emotional reactions? Does me losing at Mario Kart make me get easily annoyed. Actions like these are really difficult to track and relies on the honest of a person to make this information public. Data accuracy is the name of the game and I’d rather have a smaller set of accurate data than the big data set of non relevant information
There are a bunch of ethical issues with gathering all this data for sure. Then the question of how do you acquire the data? A lot of what we’re trying to measure is based on a physical input. A check in is not a sub conscious action it’s purposeful and if you’re not in the mood to check in then you won’t, to be honest.
Subconscious actions would provide far better mood data, imagine having your driving habits data measured. If my driving actions were measured via a black box and in turn my insurance payments are based on my previous risk assessment, well I’m all for that. How about life insurance based on your exercise and eating habits?
Anything can be measured but ultimately our mood defines a lot of our data output. I think we need to start with that.
It's All A Bet - MLOps, DevOps, AI and Synthetic Data Design 