the magical shrinking road

From: Daniel Reeves dreeves@umich.edu
Date: Tue, Jan 18, 2011 at 04:35
Subject: the magical shrinking road

Here’s an argument for Beeminder’s perhaps overcomplicated road width
adjustment algorithm. Curious if you all have thoughts on this:

[What followed has now been subsumed by this article:
http://blog.beeminder.com/roadwidth ]


http://dreev.es – search://"Daniel Reeves"
Follow the Yellow Brick Road – http://beeminder.com

From: Bethany M. Soule bsoule@gmail.com
Date: Tue, Jan 18, 2011 at 21:36
Subject: Re: the magical shrinking road

This argument is much too complicated. What’s the argument, again,
against changing the road width? Since the changing road width is what
the webservice currently provides, how about we look for compelling
arguments against using it instead of convoluted defense of it. Put
the burden of proof on the change from status quo.

On Tue, Jan 18, 2011 at 04:35, Daniel Reeves dreeves@umich.edu wrote:

Here’s an argument for Beeminder’s perhaps overcomplicated road width
adjustment algorithm. Curious if you all have thoughts on this:

Akrasia means overweighting immediate consequences. The immediate
yumminess of this pie outweighs its damage to my goal of weighing 20
pounds less in 6 months. The point of the Yellow Brick Road is to fix
that by adding immediate negative consequences to redress the
immediate positive consequences (the yumminess of the pie, the
comfiness of the couch, the wrongness on the internet
[XKCD comic on Internet arguments | Boing Boing], etc).

For weight loss, the simplest implementation of a Yellow Brick Road is
that the top edge of the road is a fixed, bright line and if you cross it you
lose. Here’s why that’s not good enough.

If that’s your contract then the right strategy is to stay a safe
distance from that bright line. A distance such that on any given day
there’s an extremely small chance of a random fluctuation pushing you
above it. That way the overall chances of ever crossing the line are,
while not “extremely small”, acceptably low. That’s the strategy
you’d like to follow, but you’re akratic [link]. Today you’re right on
that threshold of what you decided was a safe distance from the bright
you-lose line. So you ought to be super careful today. But your
short-term self, staring at today’s piece of pie, knows that the
chances of crossing the bright line are, by definition, “extremely
small”. If you eat this pie, the chances are still quite small. It’s
an acceptable risk, just this once. You eat the pie. And sure
enough, you’re only a little into that safety margin the next day.
Maybe you be good for a few days, but that nice conservative line of
safety has lost its bite. You gradually creep further and further
past it until you’re really dangerously close to the bright line.
This is motivating and you keep trying to get further below the bright
line but unless you’re quite close to it, the fear can’t compete with
the pie. So you keep too close to the edge day after day and it’s
only a matter of time before an unexpectedly large random fluctuation
pushes you over. Cue the 8-bit rendition of Chopin’s Funeral March.
[Um, that’s the standard jingle for a video game character dying; I
just spent way too long googling to figure that out.]

Why would you be so silly? Akrasia! Same reason you came to us in
the first place.

Note that this isn’t speculation. The YBR used to be like that (worse
actually; it would get wider but never narrower). We know.

What’s the solution? More immediacy! We want to take all that
gradually building risk and concentrate it on single days. How?
Roughly we want to provide a guide line that would be dangerously
close to the bright line except we make it totally safe as long as
you’re below it. (We move the bright you-lose line as necessary.)
If/when you cross that guide line we throw all the risk at you all at
once. Except, ironically, that’s so scary that it’s not so risky.
You have no choice but to panic now; you’re right up against the
bright line. The pie can’t compete with that and you scramble back
below the guide line to safety.

Notice the difference. In the simple fixed bright line version you
would panic just as much if you were ever so close to the bright line.
But you can’t ever get that close without taking on a pretty big risk
of overshooting and crossing it. So you either never get close enough
to really panic or you do and your risk of actually losing is way too
big. Compare again to the magically moving bright line where you can
and do get close enough to panic yet don’t take on the risk of
overshooting and crossing it. You’re automatically safe until panic
mode is triggered, and then you really have to panic. So you do, and
you get safe again.

Note that we’ve never described it that way so let me map it back to
how we have described it: What I called the guide line in the
previous paragraph is just the centerline, the dotted straight-line
path to your goal. And the bright line is the top edge of the road.
The bright line magically moves because the road width is equal to
your max recent deviation. And if you jump from below the centerline
to above the top of the road, well, that’s now the biggest recent
deviation, so the road widens accordingly and you’re not in fact above
the road. Hence we get the guarantee that you can’t lose tomorrow if
you’re in the right lane today. (Details have been glossed here.)


http://dreev.es – search://“Daniel Reeves”


What are you kibotzing on?


http://dreev.es – search://“Daniel Reeves”
Follow the Yellow Brick Road – http://beeminder.com

From: Daniel Reeves dreeves@umich.edu
Date: Tue, Jan 18, 2011 at 23:16
Subject: Re: the magical shrinking road

This argument is much too complicated. What’s the argument, again,
against changing the road width?

It’s that a magically changing road width is complicated (and if
you’re signing a commitment contract it should be extremely clear what
you’re agreeing to).
Patrick’s proposal is nice and simple: cross this line, you lose.

So in that sense the burden of proof is on us to justify the added
complexity of the variable road width. That’s what I tried to do
here. Not sure if anyone’s convinced though.

I think Bethany’s argument is that with just a bit of faith in
Beeminder, the status quo is just as simple: Go off this road, you
lose. (More specifically, go above the centerline and your risk of
losing goes from zero to significant, unless you immediately buckle
down and get back below the centerline. So what the road actually does
is a bit complicated, but what you actually have to do is simple.)


http://dreev.es – search://“Daniel Reeves”


http://dreev.es – search://“Daniel Reeves”
Follow the Yellow Brick Road – http://beeminder.com

From: Bethany M. Soule bsoule@gmail.com
Date: Tue, Jan 18, 2011 at 23:25
Subject: Re: the magical shrinking road

can we accomplish the best of both worlds by keeping the you-lose-line
fixed but change the ‘centerline’-s distance from it based on recent
fluctuations?

On Tue, Jan 18, 2011 at 23:16, Daniel Reeves dreeves@umich.edu wrote:

This argument is much too complicated. What’s the argument, again,
against changing the road width?

It’s that a magically changing road width is complicated (and if
you’re signing a commitment contract it should be extremely clear what
you’re agreeing to).
Patrick’s proposal is nice and simple: cross this line, you lose.

So in that sense the burden of proof is on us to justify the added
complexity of the variable road width. That’s what I tried to do
here. Not sure if anyone’s convinced though.

I think Bethany’s argument is that with just a bit of faith in
Beeminder, the status quo is just as simple: Go off this road, you
lose. (More specifically, go above the centerline and your risk of
losing goes from zero to significant, unless you immediately buckle
down and get back below the centerline. So what the road actually does
is a bit complicated, but what you actually have to do is simple.)


http://dreev.es – search://“Daniel Reeves”


What are you kibotzing on?


http://dreev.es – search://“Daniel Reeves”
Follow the Yellow Brick Road – http://beeminder.com

From: Daniel Reeves dreeves@umich.edu
Date: Tue, Jan 18, 2011 at 23:29
Subject: Re: the magical shrinking road

Good thinking, but I would say that’s out of the question because the
centerline is the straight-line path from your starting point to your
goal. It’s the idealized YBR. It would be very weird for that to
fluctuate.

On Tue, Jan 18, 2011 at 23:25, Bethany M. Soule bsoule@gmail.com wrote:

can we accomplish the best of both worlds by keeping the you-lose-line
fixed but change the ‘centerline’-s distance from it based on recent
fluctuations?

On Tue, Jan 18, 2011 at 23:16, Daniel Reeves dreeves@umich.edu wrote:

This argument is much too complicated. What’s the argument, again,
against changing the road width?

It’s that a magically changing road width is complicated (and if
you’re signing a commitment contract it should be extremely clear what
you’re agreeing to).
Patrick’s proposal is nice and simple: cross this line, you lose.

So in that sense the burden of proof is on us to justify the added
complexity of the variable road width. That’s what I tried to do
here. Not sure if anyone’s convinced though.

I think Bethany’s argument is that with just a bit of faith in
Beeminder, the status quo is just as simple: Go off this road, you
lose. (More specifically, go above the centerline and your risk of
losing goes from zero to significant, unless you immediately buckle
down and get back below the centerline. So what the road actually does
is a bit complicated, but what you actually have to do is simple.)


http://dreev.es – search://“Daniel Reeves”


What are you kibotzing on?
http://kibotzer.com


http://dreev.es – search://“Daniel Reeves”


http://dreev.es – search://“Daniel Reeves”
Follow the Yellow Brick Road – http://beeminder.com

From: Daniel Reeves dreeves@umich.edu
Date: Wed, Jan 19, 2011 at 05:17
Subject: Re: the magical shrinking road

How about this as the official rules for a weight loss road:

The road is always wide enough to guarantee that you won’t go off the
road tomorrow if you were in the right lane today. It will
automatically widen to maintain that guarantee. Additionally the road
will not shrink while you are in the wrong lane. If you’re in the
wrong lane then the edge of the road is just a fixed line: you cross
it, you lose. Otherwise, the road’s width is a magic formula based on
how much your weight fluctuates, reflecting a reasonable margin of
error around the centerline. Thanks to the “can’t lose tomorrow”
guarantee, you don’t need to know that formula. If you cross the
centerline then you’ll at worst be on the top edge of the road. You’ll
then have to lose weight at a rate equal to or faster than the
prescribed daily rate of the road until you’re back below the
centerline. In short, being in the wrong lane means a danger of a
random fluctuation tomorrow throwing you off the road. In the right
lane that is not possible.
(The beauty of this rule is that as long as you stay in the right lane
there’s no risk at all, and if you do cross into the wrong lane the
danger is sudden and significant, forcing you to buckle down
immediately.)

On Tue, Jan 18, 2011 at 23:29, Daniel Reeves dreeves@umich.edu wrote:

Good thinking, but I would say that’s out of the question because the
centerline is the straight-line path from your starting point to your
goal. It’s the idealized YBR. It would be very weird for that to
fluctuate.

On Tue, Jan 18, 2011 at 23:25, Bethany M. Soule bsoule@gmail.com wrote:

can we accomplish the best of both worlds by keeping the you-lose-line
fixed but change the ‘centerline’-s distance from it based on recent
fluctuations?

On Tue, Jan 18, 2011 at 23:16, Daniel Reeves dreeves@umich.edu wrote:

This argument is much too complicated. What’s the argument, again,
against changing the road width?

It’s that a magically changing road width is complicated (and if
you’re signing a commitment contract it should be extremely clear what
you’re agreeing to).
Patrick’s proposal is nice and simple: cross this line, you lose.

So in that sense the burden of proof is on us to justify the added
complexity of the variable road width. That’s what I tried to do
here. Not sure if anyone’s convinced though.

I think Bethany’s argument is that with just a bit of faith in
Beeminder, the status quo is just as simple: Go off this road, you
lose. (More specifically, go above the centerline and your risk of
losing goes from zero to significant, unless you immediately buckle
down and get back below the centerline. So what the road actually does
is a bit complicated, but what you actually have to do is simple.)


http://dreev.es – search://“Daniel Reeves”


What are you kibotzing on?
http://kibotzer.com


http://dreev.es – search://“Daniel Reeves”


http://dreev.es – search://“Daniel Reeves”


http://dreev.es – search://“Daniel Reeves”
Follow the Yellow Brick Road – http://beeminder.com

[I redacted the other “sounds good” replies".]

From: David Reiley david@davidreiley.com
Date: Wed, Jan 19, 2011 at 10:10
Subject: Re: the magical shrinking road

Yes, that sounds pretty understandable.

I’m wondering about more complicated bets that would help one keep
more of a safety margin.

Suppose I had a $1000 bet to lose 5 pounds in 10 weeks. Suppose I
have trouble keeping my safety buffer, for the reasons you’ve
described.

After two weeks, I should be down 1 pound. What if, for example, I
lose $100 for failing to be in the correct lane at that point? It
doesn’t kill me to be in the wrong lane - I can still recover - but it
sure would be nice if I had strong incentives to stay well within my
goal. I don’t want to lose the whole thing, because maybe some big
dining opportunity came up that I think is worth losing $100 for.
More importantly, it would be nice to have an incentive to behave in a
way that gives me a safety buffer, without having to constantly live
on the edge of the line, in danger of losing the big bet unless I am
really draconian with myself today.

This might be too complicated for Beeminder, but I wanted to propose
this for you to noodle on. It’s always bothered me that, by
bringing future incentives to the present, we still can end up in a
state of perpetual short-term misery, while trying to meet the goal,
because we didn’t leave enough of a safety buffer to be able to react
to unpredicted opportunities or challenges that come up.

David

On Jan 19, 2011, at 9:19 AM, Bethany M. Soule wrote:

Sounds good to me.

On Wed, Jan 19, 2011 at 05:17, Daniel Reeves dreeves@umich.edu wrote:

How about this as the official rules for a weight loss road:

The road is always wide enough to guarantee that you won’t go off the
road tomorrow if you were in the right lane today. It will
automatically widen to maintain that guarantee. Additionally the road
will not shrink while you are in the wrong lane. If you’re in the
wrong lane then the edge of the road is just a fixed line: you cross
it, you lose. Otherwise, the road’s width is a magic formula based on
how much your weight fluctuates, reflecting a reasonable margin of
error around the centerline. Thanks to the “can’t lose tomorrow”
guarantee, you don’t need to know that formula. If you cross the
centerline then you’ll at worst be on the top edge of the road. You’ll
then have to lose weight at a rate equal to or faster than the
prescribed daily rate of the road until you’re back below the
centerline. In short, being in the wrong lane means a danger of a
random fluctuation tomorrow throwing you off the road. In the right
lane that is not possible.
(The beauty of this rule is that as long as you stay in the right lane
there’s no risk at all, and if you do cross into the wrong lane the
danger is sudden and significant, forcing you to buckle down
immediately.)

On Tue, Jan 18, 2011 at 23:29, Daniel Reeves dreeves@umich.edu wrote:

Good thinking, but I would say that’s out of the question because the
centerline is the straight-line path from your starting point to your
goal. It’s the idealized YBR. It would be very weird for that to
fluctuate.

On Tue, Jan 18, 2011 at 23:25, Bethany M. Soule bsoule@gmail.com wrote:

can we accomplish the best of both worlds by keeping the you-lose-line
fixed but change the ‘centerline’-s distance from it based on recent
fluctuations?

On Tue, Jan 18, 2011 at 23:16, Daniel Reeves dreeves@umich.edu wrote:

This argument is much too complicated. What’s the argument, again,
against changing the road width?

It’s that a magically changing road width is complicated (and if
you’re signing a commitment contract it should be extremely clear what
you’re agreeing to).
Patrick’s proposal is nice and simple: cross this line, you lose.

So in that sense the burden of proof is on us to justify the added
complexity of the variable road width. That’s what I tried to do
here. Not sure if anyone’s convinced though.

I think Bethany’s argument is that with just a bit of faith in
Beeminder, the status quo is just as simple: Go off this road, you
lose. (More specifically, go above the centerline and your risk of
losing goes from zero to significant, unless you immediately buckle
down and get back below the centerline. So what the road actually does
is a bit complicated, but what you actually have to do is simple.)


http://dreev.es – search://“Daniel Reeves”


What are you kibotzing on?
http://kibotzer.com


http://dreev.es – search://“Daniel Reeves”


http://dreev.es – search://“Daniel Reeves”


What are you kibotzing on?
http://kibotzer.com


My home page: http://www.davidreiley.com


http://dreev.es – search://“Daniel Reeves”
Follow the Yellow Brick Road – http://beeminder.com

From: Patrick Jordan pjordan@reasoningtech.com
Date: Wed, Jan 19, 2011 at 11:11
Subject: Re: the magical shrinking road

I still don’t see what the magic part is adding or how it is coming into play.

In terms of what David said, would the following work:

You have a total bet of X for achieving your (global) goal. The total
bet is broken down into weekly bets x, where you measure daily to see
if you have violated your weekly contract. Each week you have to
achieve a specific goal that is on a linear path to your global goal
plus some error tolerance E, starting from the prior week’s endpoint.
The rate at which you have to loose automatically adjusts each reset,
but the system automatically converges to the global value.

Properties:

  • If you are losing weight quicker than the contracted rate, each
    week’s subsequent rate is easier to achieve.
  • If you are losing weight slower than the initial contracted rate,
    each week’s rate is harder to achieve until you get back to the
    initial contracted rate.
  • E encapsulates the risk from systematic error.

Patrick Jordan
http://www.patrickrjordan.com
pjordan@reasoningtech.com

On Jan 19, 2011, at 8:10 AM, David Reiley wrote:

Yes, that sounds pretty understandable.

I’m wondering about more complicated bets that would help one keep more of a safety margin.

Suppose I had a $1000 bet to lose 5 pounds in 10 weeks. Suppose I have trouble keeping my safety buffer, for the reasons you’ve described.

After two weeks, I should be down 1 pound. What if, for example, I lose $100 for failing to be in the correct lane at that point? It doesn’t kill me to be in the wrong lane - I can still recover - but it sure would be nice if I had strong incentives to stay well within my goal. I don’t want to lose the whole thing, because maybe some big dining opportunity came up that I think is worth losing $100 for. More importantly, it would be nice to have an incentive to behave in a way that gives me a safety buffer, without having to constantly live on the edge of the line, in danger of losing the big bet unless I am really draconian with myself today.

This might be too complicated for Beeminder, but I wanted to propose this for you to noodle on. It’s always bothered me that, by bringing future incentives to the present, we still can end up in a state of perpetual short-term misery, while trying to meet the goal, because we didn’t leave enough of a safety buffer to be able to react to unpredicted opportunities or challenges that come up.

David

On Jan 19, 2011, at 9:19 AM, Bethany M. Soule wrote:

Sounds good to me.

On Wed, Jan 19, 2011 at 05:17, Daniel Reeves dreeves@umich.edu wrote:

How about this as the official rules for a weight loss road:

The road is always wide enough to guarantee that you won’t go off the
road tomorrow if you were in the right lane today. It will
automatically widen to maintain that guarantee. Additionally the road
will not shrink while you are in the wrong lane. If you’re in the
wrong lane then the edge of the road is just a fixed line: you cross
it, you lose. Otherwise, the road’s width is a magic formula based on
how much your weight fluctuates, reflecting a reasonable margin of
error around the centerline. Thanks to the “can’t lose tomorrow”
guarantee, you don’t need to know that formula. If you cross the
centerline then you’ll at worst be on the top edge of the road. You’ll
then have to lose weight at a rate equal to or faster than the
prescribed daily rate of the road until you’re back below the
centerline. In short, being in the wrong lane means a danger of a
random fluctuation tomorrow throwing you off the road. In the right
lane that is not possible.
(The beauty of this rule is that as long as you stay in the right lane
there’s no risk at all, and if you do cross into the wrong lane the
danger is sudden and significant, forcing you to buckle down
immediately.)

On Tue, Jan 18, 2011 at 23:29, Daniel Reeves dreeves@umich.edu wrote:

Good thinking, but I would say that’s out of the question because the
centerline is the straight-line path from your starting point to your
goal. It’s the idealized YBR. It would be very weird for that to
fluctuate.

On Tue, Jan 18, 2011 at 23:25, Bethany M. Soule bsoule@gmail.com wrote:

can we accomplish the best of both worlds by keeping the you-lose-line
fixed but change the ‘centerline’-s distance from it based on recent
fluctuations?

On Tue, Jan 18, 2011 at 23:16, Daniel Reeves dreeves@umich.edu wrote:

This argument is much too complicated. What’s the argument, again,
against changing the road width?

It’s that a magically changing road width is complicated (and if
you’re signing a commitment contract it should be extremely clear what
you’re agreeing to).
Patrick’s proposal is nice and simple: cross this line, you lose.

So in that sense the burden of proof is on us to justify the added
complexity of the variable road width. That’s what I tried to do
here. Not sure if anyone’s convinced though.

I think Bethany’s argument is that with just a bit of faith in
Beeminder, the status quo is just as simple: Go off this road, you
lose. (More specifically, go above the centerline and your risk of
losing goes from zero to significant, unless you immediately buckle
down and get back below the centerline. So what the road actually does
is a bit complicated, but what you actually have to do is simple.)


http://dreev.es – search://“Daniel Reeves”


What are you kibotzing on?
http://kibotzer.com


http://dreev.es – search://“Daniel Reeves”


http://dreev.es – search://“Daniel Reeves”


What are you kibotzing on?
http://kibotzer.com


My home page: http://www.davidreiley.com

From: Patrick Jordan pjordan@reasoningtech.com
Date: Wed, Jan 19, 2011 at 11:46
Subject: Re: the magical shrinking road

After a half-hour’s reflection, how about the following:

Since we have no principled way of comparing road policies by way of
thought experiment, we setup controlled experiments for different
policies and measure their effectiveness.

Policy constraint: For an given day, a policy will state the weight
the user has to achieve by tomorrow to not violate the policy.

I believe that is possible for each of the policies we have discussed.


Patrick Jordan
http://www.patrickrjordan.com
pjordan@reasoningtech.com

On Jan 19, 2011, at 9:11 AM, Patrick Jordan wrote:

I still don’t see what the magic part is adding or how it is coming into play.

In terms of what David said, would the following work:

You have a total bet of X for achieving your (global) goal. The total bet is broken down into weekly bets x, where you measure daily to see if you have violated your weekly contract. Each week you have to achieve a specific goal that is on a linear path to your global goal plus some error tolerance E, starting from the prior week’s endpoint. The rate at which you have to loose automatically adjusts each reset, but the system automatically converges to the global value.

Properties:

  • If you are loosing weight quicker than the contracted rate, each week’s subsequent rate is easier to achieve.
  • If you are loosing weight slower than the initial contracted rate, each week’s rate is harder to achieve until you get back to the initial contracted rate.
  • E encapsulates the risk from systematic error.

Patrick Jordan
http://www.patrickrjordan.com
pjordan@reasoningtech.com

On Jan 19, 2011, at 8:10 AM, David Reiley wrote:

Yes, that sounds pretty understandable.

I’m wondering about more complicated bets that would help one keep more of a safety margin.

Suppose I had a $1000 bet to lose 5 pounds in 10 weeks. Suppose I have trouble keeping my safety buffer, for the reasons you’ve described.

After two weeks, I should be down 1 pound. What if, for example, I lose $100 for failing to be in the correct lane at that point? It doesn’t kill me to be in the wrong lane - I can still recover - but it sure would be nice if I had strong incentives to stay well within my goal. I don’t want to lose the whole thing, because maybe some big dining opportunity came up that I think is worth losing $100 for. More importantly, it would be nice to have an incentive to behave in a way that gives me a safety buffer, without having to constantly live on the edge of the line, in danger of losing the big bet unless I am really draconian with myself today.

This might be too complicated for Beeminder, but I wanted to propose this for you to noodle on. It’s always bothered me that, by bringing future incentives to the present, we still can end up in a state of perpetual short-term misery, while trying to meet the goal, because we didn’t leave enough of a safety buffer to be able to react to unpredicted opportunities or challenges that come up.

David

On Jan 19, 2011, at 9:19 AM, Bethany M. Soule wrote:

Sounds good to me.

On Wed, Jan 19, 2011 at 05:17, Daniel Reeves dreeves@umich.edu wrote:

How about this as the official rules for a weight loss road:

The road is always wide enough to guarantee that you won’t go off the
road tomorrow if you were in the right lane today. It will
automatically widen to maintain that guarantee. Additionally the road
will not shrink while you are in the wrong lane. If you’re in the
wrong lane then the edge of the road is just a fixed line: you cross
it, you lose. Otherwise, the road’s width is a magic formula based on
how much your weight fluctuates, reflecting a reasonable margin of
error around the centerline. Thanks to the “can’t lose tomorrow”
guarantee, you don’t need to know that formula. If you cross the
centerline then you’ll at worst be on the top edge of the road. You’ll
then have to lose weight at a rate equal to or faster than the
prescribed daily rate of the road until you’re back below the
centerline. In short, being in the wrong lane means a danger of a
random fluctuation tomorrow throwing you off the road. In the right
lane that is not possible.
(The beauty of this rule is that as long as you stay in the right lane
there’s no risk at all, and if you do cross into the wrong lane the
danger is sudden and significant, forcing you to buckle down
immediately.)

On Tue, Jan 18, 2011 at 23:29, Daniel Reeves dreeves@umich.edu wrote:

Good thinking, but I would say that’s out of the question because the
centerline is the straight-line path from your starting point to your
goal. It’s the idealized YBR. It would be very weird for that to
fluctuate.

On Tue, Jan 18, 2011 at 23:25, Bethany M. Soule bsoule@gmail.com wrote:

can we accomplish the best of both worlds by keeping the you-lose-line
fixed but change the ‘centerline’-s distance from it based on recent
fluctuations?

On Tue, Jan 18, 2011 at 23:16, Daniel Reeves dreeves@umich.edu wrote:

This argument is much too complicated. What’s the argument, again,
against changing the road width?

It’s that a magically changing road width is complicated (and if
you’re signing a commitment contract it should be extremely clear what
you’re agreeing to).
Patrick’s proposal is nice and simple: cross this line, you lose.

So in that sense the burden of proof is on us to justify the added
complexity of the variable road width. That’s what I tried to do
here. Not sure if anyone’s convinced though.

I think Bethany’s argument is that with just a bit of faith in
Beeminder, the status quo is just as simple: Go off this road, you
lose. (More specifically, go above the centerline and your risk of
losing goes from zero to significant, unless you immediately buckle
down and get back below the centerline. So what the road actually does
is a bit complicated, but what you actually have to do is simple.)


http://dreev.es – search://“Daniel Reeves”


What are you kibotzing on?
http://kibotzer.com


http://dreev.es – search://“Daniel Reeves”


http://dreev.es – search://“Daniel Reeves”


What are you kibotzing on?
http://kibotzer.com


My home page: http://www.davidreiley.com

From: Daniel Reeves dreeves@kibotzer.com
Date: Wed, Jan 19, 2011 at 11:50
Subject: Re: the magical shrinking road

I still don’t see what the magic part is adding or how it is coming into play.

The magic part has (quite intentionally) little impact now. (Compare
to a magic formula that always collapsed the road to zero width
whenever it could. That wouldn’t really be any harder.) But you
still want some way to decide the width of the road when you’re in the
right lane, to give some idea of how much leeway you have. One way to
think of the magic formula is: it’s our guess as to where the fixed
you-lose line will be when/if you go above the centerline and a
you-lose line is instantiated.

(I have to think more about your/David’s proposal of intermediate penalties.)

From: David Reiley david@davidreiley.com
Date: Wed, Jan 19, 2011 at 13:03
Subject: Re: the magical shrinking road

I love the idea. Right up my career alley.

Of course, it might be that different policies are better for
different people. But to begin, it might be sufficient to ask what
policy gives the most benefit to the most people. Later, with lots
of data, one could tell whether some types of people respond better to
different kinds of policies.

David

On Jan 19, 2011, at 12:46 PM, Patrick Jordan wrote:

After a half-hour’s reflection, how about the following:

Since we have no principled way of comparing road policies by way of thought experiment, we setup controlled experiments for different policies and measure their effectiveness.

Policy constraint: For an given day, a policy will state the weight the user has to achieve by tomorrow to not violate the policy.

I believe that is possible for each of the policies we have discussed.


Patrick Jordan
http://www.patrickrjordan.com
pjordan@reasoningtech.com

On Jan 19, 2011, at 9:11 AM, Patrick Jordan wrote:

I still don’t see what the magic part is adding or how it is coming into play.

In terms of what David said, would the following work:

You have a total bet of X for achieving your (global) goal. The total bet is broken down into weekly bets x, where you measure daily to see if you have violated your weekly contract. Each week you have to achieve a specific goal that is on a linear path to your global goal plus some error tolerance E, starting from the prior week’s endpoint. The rate at which you have to loose automatically adjusts each reset, but the system automatically converges to the global value.

Properties:

  • If you are loosing weight quicker than the contracted rate, each week’s subsequent rate is easier to achieve.
  • If you are loosing weight slower than the initial contracted rate, each week’s rate is harder to achieve until you get back to the initial contracted rate.
  • E encapsulates the risk from systematic error.

Patrick Jordan
http://www.patrickrjordan.com
pjordan@reasoningtech.com

On Jan 19, 2011, at 8:10 AM, David Reiley wrote:

Yes, that sounds pretty understandable.

I’m wondering about more complicated bets that would help one keep more of a safety margin.

Suppose I had a $1000 bet to lose 5 pounds in 10 weeks. Suppose I have trouble keeping my safety buffer, for the reasons you’ve described.

After two weeks, I should be down 1 pound. What if, for example, I lose $100 for failing to be in the correct lane at that point? It doesn’t kill me to be in the wrong lane - I can still recover - but it sure would be nice if I had strong incentives to stay well within my goal. I don’t want to lose the whole thing, because maybe some big dining opportunity came up that I think is worth losing $100 for. More importantly, it would be nice to have an incentive to behave in a way that gives me a safety buffer, without having to constantly live on the edge of the line, in danger of losing the big bet unless I am really draconian with myself today.

This might be too complicated for Beeminder, but I wanted to propose this for you to noodle on. It’s always bothered me that, by bringing future incentives to the present, we still can end up in a state of perpetual short-term misery, while trying to meet the goal, because we didn’t leave enough of a safety buffer to be able to react to unpredicted opportunities or challenges that come up.

David

On Jan 19, 2011, at 9:19 AM, Bethany M. Soule wrote:

Sounds good to me.

On Wed, Jan 19, 2011 at 05:17, Daniel Reeves dreeves@umich.edu wrote:

How about this as the official rules for a weight loss road:

The road is always wide enough to guarantee that you won’t go off the
road tomorrow if you were in the right lane today. It will
automatically widen to maintain that guarantee. Additionally the road
will not shrink while you are in the wrong lane. If you’re in the
wrong lane then the edge of the road is just a fixed line: you cross
it, you lose. Otherwise, the road’s width is a magic formula based on
how much your weight fluctuates, reflecting a reasonable margin of
error around the centerline. Thanks to the “can’t lose tomorrow”
guarantee, you don’t need to know that formula. If you cross the
centerline then you’ll at worst be on the top edge of the road. You’ll
then have to lose weight at a rate equal to or faster than the
prescribed daily rate of the road until you’re back below the
centerline. In short, being in the wrong lane means a danger of a
random fluctuation tomorrow throwing you off the road. In the right
lane that is not possible.
(The beauty of this rule is that as long as you stay in the right lane
there’s no risk at all, and if you do cross into the wrong lane the
danger is sudden and significant, forcing you to buckle down
immediately.)

On Tue, Jan 18, 2011 at 23:29, Daniel Reeves dreeves@umich.edu wrote:

Good thinking, but I would say that’s out of the question because the
centerline is the straight-line path from your starting point to your
goal. It’s the idealized YBR. It would be very weird for that to
fluctuate.

On Tue, Jan 18, 2011 at 23:25, Bethany M. Soule bsoule@gmail.com wrote:

can we accomplish the best of both worlds by keeping the you-lose-line
fixed but change the ‘centerline’-s distance from it based on recent
fluctuations?

On Tue, Jan 18, 2011 at 23:16, Daniel Reeves dreeves@umich.edu wrote:

This argument is much too complicated. What’s the argument, again,
against changing the road width?

It’s that a magically changing road width is complicated (and if
you’re signing a commitment contract it should be extremely clear what
you’re agreeing to).
Patrick’s proposal is nice and simple: cross this line, you lose.

So in that sense the burden of proof is on us to justify the added
complexity of the variable road width. That’s what I tried to do
here. Not sure if anyone’s convinced though.

I think Bethany’s argument is that with just a bit of faith in
Beeminder, the status quo is just as simple: Go off this road, you
lose. (More specifically, go above the centerline and your risk of
losing goes from zero to significant, unless you immediately buckle
down and get back below the centerline. So what the road actually does
is a bit complicated, but what you actually have to do is simple.)


http://dreev.es – search://“Daniel Reeves”


What are you kibotzing on?
http://kibotzer.com


http://dreev.es – search://“Daniel Reeves”


http://dreev.es – search://“Daniel Reeves”


What are you kibotzing on?
http://kibotzer.com


My home page: http://www.davidreiley.com


My home page: http://www.davidreiley.com


http://dreev.es – search://“Daniel Reeves”
Follow the Yellow Brick Road – http://beeminder.com

From: Daniel Reeves dreeves@umich.edu
Date: Thu, Jan 20, 2011 at 04:12
Subject: Re: the magical shrinking road

I think I’m opposed to intermediate penalties like this (though it’s
certainly worth experimenting with!).

To restate the problem you’ve identified: By self-binding you’ve
sacrificed flexibility and that’s sometimes costly. Of course that’s
kind of the point: you were abusing the flexibility and so you needed
to cut yourself off. But it wasn’t all abuse. So can we have the
best of both worlds: flexibility only when it’s really warranted? A
freebie day when you win a free meal at a five-star restaurant but not
when pie in the fridge is calling to you.

One way is to stipulate exemption clauses in the commitment contract.
Of course you won’t anticipate everything.
Another way is to specify a generic budget of freebie days, but in our
experience this doesn’t work well. If you’re so akratic (as many of
us are!) that you can’t keep a reasonable safety buffer then you end
up just using up the freebie days early. (I could say a lot more
about why attempts to make contracts kinder/gentler tend to backfire.)

So what about your proposal that you risk a big chunk of money but you
lose it in pieces each time you deviate? That’s stickK’s approach and
it’s not crazy (and, again, I agree we should try it, later). My
objection to it is that as an akratic it’s all too easy to shrug off
paying the small penalty “just this once” and end up coughing up an
unreasonable amount in total (Ayres gives examples of that happening
in his book!). If the small penalty of $100 is big enough to prevent
that problem then you might as well just use $100 as the total
penalty, Beeminder style. (The one thing missing there is that you
want to automatically restart on a new road – I think Beeminder
should support that.)

Final note: the free meal at the five-star restaurant example I
believe is not a problem in practice with the auto-widening road.
When the stakes are high people seem to be incentivized to mostly stay
in the right lane and that means you’re usually ready for that
five-star restaurant opportunity.

On Wed, Jan 19, 2011 at 10:10, David Reiley david@davidreiley.com wrote:

Yes, that sounds pretty understandable.

I’m wondering about more complicated bets that would help one keep more of a safety margin.

Suppose I had a $1000 bet to lose 5 pounds in 10 weeks. Suppose I have trouble keeping my safety buffer, for the reasons you’ve described.

After two weeks, I should be down 1 pound. What if, for example, I lose $100 for failing to be in the correct lane at that point? It doesn’t kill me to be in the wrong lane - I can still recover - but it sure would be nice if I had strong incentives to stay well within my goal. I don’t want to lose the whole thing, because maybe some big dining opportunity came up that I think is worth losing $100 for. More importantly, it would be nice to have an incentive to behave in a way that gives me a safety buffer, without having to constantly live on the edge of the line, in danger of losing the big bet unless I am really draconian with myself today.

This might be too complicated for Beeminder, but I wanted to propose this for you to noodle on. It’s always bothered me that, by bringing future incentives to the present, we still can end up in a state of perpetual short-term misery, while trying to meet the goal, because we didn’t leave enough of a safety buffer to be able to react to unpredicted opportunities or challenges that come up.

David

On Jan 19, 2011, at 9:19 AM, Bethany M. Soule wrote:

Sounds good to me.

On Wed, Jan 19, 2011 at 05:17, Daniel Reeves dreeves@umich.edu wrote:

How about this as the official rules for a weight loss road:

The road is always wide enough to guarantee that you won’t go off the
road tomorrow if you were in the right lane today. It will
automatically widen to maintain that guarantee. Additionally the road
will not shrink while you are in the wrong lane. If you’re in the
wrong lane then the edge of the road is just a fixed line: you cross
it, you lose. Otherwise, the road’s width is a magic formula based on
how much your weight fluctuates, reflecting a reasonable margin of
error around the centerline. Thanks to the “can’t lose tomorrow”
guarantee, you don’t need to know that formula. If you cross the
centerline then you’ll at worst be on the top edge of the road. You’ll
then have to lose weight at a rate equal to or faster than the
prescribed daily rate of the road until you’re back below the
centerline. In short, being in the wrong lane means a danger of a
random fluctuation tomorrow throwing you off the road. In the right
lane that is not possible.
(The beauty of this rule is that as long as you stay in the right lane
there’s no risk at all, and if you do cross into the wrong lane the
danger is sudden and significant, forcing you to buckle down
immediately.)

On Tue, Jan 18, 2011 at 23:29, Daniel Reeves dreeves@umich.edu wrote:

Good thinking, but I would say that’s out of the question because the
centerline is the straight-line path from your starting point to your
goal. It’s the idealized YBR. It would be very weird for that to
fluctuate.

On Tue, Jan 18, 2011 at 23:25, Bethany M. Soule bsoule@gmail.com wrote:

can we accomplish the best of both worlds by keeping the you-lose-line
fixed but change the ‘centerline’-s distance from it based on recent
fluctuations?

On Tue, Jan 18, 2011 at 23:16, Daniel Reeves dreeves@umich.edu wrote:

This argument is much too complicated. What’s the argument, again,
against changing the road width?

It’s that a magically changing road width is complicated (and if
you’re signing a commitment contract it should be extremely clear what
you’re agreeing to).
Patrick’s proposal is nice and simple: cross this line, you lose.

So in that sense the burden of proof is on us to justify the added
complexity of the variable road width. That’s what I tried to do
here. Not sure if anyone’s convinced though.

I think Bethany’s argument is that with just a bit of faith in
Beeminder, the status quo is just as simple: Go off this road, you
lose. (More specifically, go above the centerline and your risk of
losing goes from zero to significant, unless you immediately buckle
down and get back below the centerline. So what the road actually does
is a bit complicated, but what you actually have to do is simple.)


http://dreev.es – search://“Daniel Reeves”


What are you kibotzing on?
http://kibotzer.com


http://dreev.es – search://“Daniel Reeves”


http://dreev.es – search://“Daniel Reeves”


What are you kibotzing on?
http://kibotzer.com


My home page: http://www.davidreiley.com


http://dreev.es – search://“Daniel Reeves”


http://dreev.es – search://“Daniel Reeves”
Follow the Yellow Brick Road – http://beeminder.com