On the other hand the existence doesn’t mean that a new attempt at the same problem is worthless. I think it’s very valuable to have multiple people try to solve the problem.
To me it seems like a much more interesting project than having another go at writing an app to parse an online forum.
There are few people thinking in depth about designing apps to teach people to be calibrated.
The fact that you have a smartphone also allows additional questions:
You can ask calibration questions such as:
Did John or Joe send you more emails in the last year?
Is the air pressure more or less than X?
Is the temperature of the smart phone battery more or less than X?
Does this arrow point more North or more South?”
Is the distance between your work location and where you are at the moment more or less than X?
Is the distance between your home location and where you are at the moment more or less than X?
Is the distance between where John lives and where you are at the moment more or less than X?
What was the average speed at which you where traveling in the last minute (if you sit in public transportation)
Is the average pitch of the background noise over the last minute more or less than X?
Is the longest email that you received in the past week more or less than X characters long?
What’s the chance that you will get a call today?”
Is the average of beeminder value X that you tracked over the last week (month) more or less than X?
All those questions are more interesting then whether postmaster general X served before or after postmaster general Y or the boiling temperatures of various metals.
Building an app around the issue might be more complicated than simply providing an new interface for LessWrong, but the payoff for getting Credence training right is also so much higher.
Even if you simply focus on building a beeminder history credence game that might not be too complicated but really useful.
Too me it feels like a waste to have valuable development resources wasted on building a Lesswrong app when there are much more valuable projects.
Simple version: You provide your own predictions, and state your credence. Later you say whether you were right or wrong. The app displays statistics of your calibration.
This is simple in essence, but there will be many design decisions, and many little details that can make the UI better. For example, I guess you should choose the credence from, say, 50%, 60%, 70%, 80%, 90%, 95%, and 99%, instead of typing your own value, because this way it will be easier to make statistics. Also, choosing one option is easier than typing two digits, although most of the work will be typing the questions. It should be possible to edit the text later (noticing a typo too late would drive me crazy). The app should also remember the date each question was entered, so it can give you statistics like: how well calibrated you are in the last 30 days (compared with the previous 30 days).
Maybe the data should be stored online, so you can edit them both from the mobile and from the PC. Although, I would prefer if the application works offline, too. These are two contradictory demands, so you have to find a solution. Perhaps each user should choose in settings whether their data should be kept in the mobile or on the web? And perhaps allow to change this setting later, and the data will be copied? Or maybe even keeping only the recent data in the mobile, and the full archive online? There are many decisions here.
A nice function would be to save some work typing repeated questions. For example, if I want to make a bet every morning “will I exercise today?”, there should be an option to repeat one of the recent questions with current date. (By the way, if you always display the date along the question, you can write things like “today” or “this month” without having to always write the specific date.)
A more advanced version (don’t do this as the first version; remember the planning fallacy!) would allow some kind of “multiplayer”. You could add friends, and offer to share some bets with your friends. Anyone can create a question and offer it to other people; they can accept (by writing their credence) or reject it. Then there would be a summary comparing the members of the group.
Again, here are many design choices and UI improvements. How specifically will you add friends? Will you also have groups of friends, so you share some questions only with some groups? Who can answer the multiplayer question: the person who wrote it, anyone, or the person who wrote it chooses one of the former options?
Integrate the whole thing with Facebook, especially the multiplayer version? That could make the app wildly popular! (But I heard that the Facebook API is less than friendly.)
There already is one for Android.
I wasn’t aware of the Android app.
On the other hand the existence doesn’t mean that a new attempt at the same problem is worthless. I think it’s very valuable to have multiple people try to solve the problem.
To me it seems like a much more interesting project than having another go at writing an app to parse an online forum. There are few people thinking in depth about designing apps to teach people to be calibrated.
The fact that you have a smartphone also allows additional questions:
You can ask calibration questions such as:
Did John or Joe send you more emails in the last year?
Is the air pressure more or less than X?
Is the temperature of the smart phone battery more or less than X?
Does this arrow point more North or more South?”
Is the distance between your work location and where you are at the moment more or less than X?
Is the distance between your home location and where you are at the moment more or less than X?
Is the distance between where John lives and where you are at the moment more or less than X?
What was the average speed at which you where traveling in the last minute (if you sit in public transportation)
Is the average pitch of the background noise over the last minute more or less than X?
Is the longest email that you received in the past week more or less than X characters long?
What’s the chance that you will get a call today?”
Is the average of beeminder value X that you tracked over the last week (month) more or less than X?
All those questions are more interesting then whether postmaster general X served before or after postmaster general Y or the boiling temperatures of various metals. Building an app around the issue might be more complicated than simply providing an new interface for LessWrong, but the payoff for getting Credence training right is also so much higher.
Even if you simply focus on building a beeminder history credence game that might not be too complicated but really useful. Too me it feels like a waste to have valuable development resources wasted on building a Lesswrong app when there are much more valuable projects.
Just wanted to say: thanks for the ideas!
A personal prediction book?
Simple version: You provide your own predictions, and state your credence. Later you say whether you were right or wrong. The app displays statistics of your calibration.
This is simple in essence, but there will be many design decisions, and many little details that can make the UI better. For example, I guess you should choose the credence from, say, 50%, 60%, 70%, 80%, 90%, 95%, and 99%, instead of typing your own value, because this way it will be easier to make statistics. Also, choosing one option is easier than typing two digits, although most of the work will be typing the questions. It should be possible to edit the text later (noticing a typo too late would drive me crazy). The app should also remember the date each question was entered, so it can give you statistics like: how well calibrated you are in the last 30 days (compared with the previous 30 days).
Maybe the data should be stored online, so you can edit them both from the mobile and from the PC. Although, I would prefer if the application works offline, too. These are two contradictory demands, so you have to find a solution. Perhaps each user should choose in settings whether their data should be kept in the mobile or on the web? And perhaps allow to change this setting later, and the data will be copied? Or maybe even keeping only the recent data in the mobile, and the full archive online? There are many decisions here.
A nice function would be to save some work typing repeated questions. For example, if I want to make a bet every morning “will I exercise today?”, there should be an option to repeat one of the recent questions with current date. (By the way, if you always display the date along the question, you can write things like “today” or “this month” without having to always write the specific date.)
A more advanced version (don’t do this as the first version; remember the planning fallacy!) would allow some kind of “multiplayer”. You could add friends, and offer to share some bets with your friends. Anyone can create a question and offer it to other people; they can accept (by writing their credence) or reject it. Then there would be a summary comparing the members of the group.
Again, here are many design choices and UI improvements. How specifically will you add friends? Will you also have groups of friends, so you share some questions only with some groups? Who can answer the multiplayer question: the person who wrote it, anyone, or the person who wrote it chooses one of the former options?
Integrate the whole thing with Facebook, especially the multiplayer version? That could make the app wildly popular! (But I heard that the Facebook API is less than friendly.)