Ask HN: How AI/ML can be applied to improve our phone usage habits?
7 comments
How do you think ad targeting works? First google and friends build a model of your digital habits and then they figure out the best ways to intervene in the digital space with ads that you will click on. Same with Facebook. These companies make billions every day and that's because they are very effective at using AI (linear regression) to modify user behavior on a mass scale and increase revenue for the advertisers.
Apologies for misleading title of my question, I was curious about "improving" our phone usage habits not using it for the purpose of getting more attention. Something like reducing screen time.
You can use the same principles that Google and friends use to intervene on other behaviors. First you need data to make a model and then you use the model to optimize some metric. In this case it seems like you want to reduce screen time so the intervention in this case would be to lock out applications that seem like unproductive time sinks. Alternatively, you can look into developing better meta-cognitive skills and you don't need extra software for that other than spending some time every day meditating.
You don't even need AI for this. Just track application use and then gradually increase the lockout time for the top 3 applications.
You don't even need AI for this. Just track application use and then gradually increase the lockout time for the top 3 applications.
Then a proper question is "what metric to select for optimization?"
As for just using plain software, yes it is an option indeed. But when we are using our devices, we produce so many data points and they are mostly used for hooking our attention through models. Don't you think that there is there so little reverse targeted application of the models?
As for just using plain software, yes it is an option indeed. But when we are using our devices, we produce so many data points and they are mostly used for hooking our attention through models. Don't you think that there is there so little reverse targeted application of the models?
Getting people to waste time is profitable so that's why AI is used to waste people's time and get them to buy stuff. It's just how the world economy works. If people stop consuming and buying stuff they don't need then the system stops working.
Using what data as input?
And what do you mean by MSE of around 3 minutes? Do you mean root mean squared error?
And what do you mean by MSE of around 3 minutes? Do you mean root mean squared error?
I used my personal usage data in the format
timestamp_of_session, duration_in_seconds
Yes, sorry for lack of details I meant root MSE
timestamp_of_session, duration_in_seconds
Yes, sorry for lack of details I meant root MSE
I've been deep-diving into an intriguing question while working on my side project and during my studies.
I experimented with predicting the length of my future phone usage sessions (from unlock to lock), and the outcomes have been pretty remarkable.
I got my usage data in format
timestamp_of_session, duration_in_seconds
By applying a straightforward linear regression model, I achieved a Mean Squared Error of just around 3 minutes. Isn't that fascinating?
I'm really excited to hear your thoughts and insights on this! Thanks in advance for your input!