Created
August 1, 2020 15:48
-
-
Save lukavdplas/1746e982f73b89628c8b23710ef56335 to your computer and use it in GitHub Desktop.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
### A Pluto.jl notebook ### | |
# v0.11.0 | |
using Markdown | |
using InteractiveUtils | |
# This Pluto notebook uses @bind for interactivity. When running this notebook outside of Pluto, the following 'mock version' of @bind gives bound variables a default value (instead of an error). | |
macro bind(def, element) | |
quote | |
local el = $(esc(element)) | |
global $(esc(def)) = Core.applicable(Base.get, el) ? Base.get(el) : missing | |
el | |
end | |
end | |
# ╔═╡ e6fab282-cd9d-11ea-2efb-156e0cee2c25 | |
md""" | |
## Prior distribution | |
Let's start by defining the prior height distribution in the population. | |
""" | |
# ╔═╡ d48144f8-cd9e-11ea-0daf-81c5bda3ba7c | |
md""" | |
## Literal listener | |
The literal listener assumes that a person can be called _tall_ if their height is above a fixed threshold. | |
If they know nothing about a person, their belief about the person's height follows the regular height distribution. If a person is described as _tall_, the listener can update their belief about the height of that person. | |
Specifically, they assume that there is a 0 probability that the person has a height below the threshold. The probability for a height x above the threshold must be normalised as a result. | |
""" | |
# ╔═╡ df623e2e-cda1-11ea-3fea-6d9947eda4c7 | |
md""" | |
## Expected success and utility | |
We move on to a speaker, who is describing a person that they know the height of. They can choose to describe that person as tall, or to say nothing about their height. | |
Whichever option they choose, the _success_ of their action is measured by the _belief_ of the listener. Specifically, when the speaker describes a person of height `x`, the success of the action is measured by the listener's degree of belief that the height of the person is `x`. | |
In this case, the listener is a literal listener. Their belief is given in the function above. The speaker will describe a person as _tall_ if their height is above the threshold, and will say nothing if their height is below the threshold. | |
This gives a success value for each potential height of the person described (from 0 to 250 cm). However, not all of these datapoints are equally likely, so we will weigh the success of describing a person with height `x` by the probability that a person would have height `x`. | |
""" | |
# ╔═╡ c3c288a0-cdaf-11ea-2171-edff8c53752d | |
md""" | |
We will calculate the threshold probability for each of the scale points right now, which will make calculations more efficient. | |
(Thanks to the reactive notebook, this will update automatically.) | |
""" | |
# ╔═╡ 1e810430-cd9f-11ea-0905-0936032f0267 | |
md""" | |
## Imports | |
Don't mind this | |
""" | |
# ╔═╡ 5e827892-cd9d-11ea-1cda-1d8fc9dbd8d2 | |
using PlutoUI | |
# ╔═╡ 9f2ebd74-cd9d-11ea-2f13-cfe68a07c244 | |
using Distributions | |
# ╔═╡ 04153814-cd9f-11ea-370a-85097498239b | |
using Plots | |
# ╔═╡ 277190c8-cd9d-11ea-0aee-7952ef163cb2 | |
md""" | |
Mean height: | |
50 $(@bind height_mean Slider(50:200)) 200 | |
""" | |
# ╔═╡ 7d1c8156-cd9d-11ea-1e94-816f1ab3173d | |
md""" | |
Standard deviation: | |
0 $(@bind height_sd Slider(0:100)) 100 | |
""" | |
# ╔═╡ 9bd38234-cd9d-11ea-1747-09a10ca39421 | |
height_dist = Normal(height_mean, height_sd) | |
# ╔═╡ a64fe396-cd9e-11ea-3a4f-97ef4f9b29b3 | |
function height_density(x) | |
pdf(height_dist, x) | |
end | |
# ╔═╡ be288928-cd9e-11ea-23fb-1500bd9d535f | |
function height_cumulative(x) | |
cdf(height_dist, x) | |
end | |
# ╔═╡ 47535af4-cd9f-11ea-0542-8b04b08f36a7 | |
plot(0:250, height_density, | |
legend = false, title = "Prior distribution of height", xlabel = "height", ylabel = "probability") | |
# ╔═╡ 02c0fd3c-cd9e-11ea-09ac-351343aa0c98 | |
function literal_listener(x, threshold, densityf, cumulativef) | |
if x >= threshold | |
densityf(x)/(1-cumulativef(threshold)) | |
else | |
0 | |
end | |
end | |
# ╔═╡ a642e89c-cda0-11ea-2466-a7f173352f99 | |
md""" | |
Threshold: | |
0 $(@bind example_threshold Slider(50:250)) 250 | |
""" | |
# ╔═╡ 82e11130-cda0-11ea-3099-db1268bdf72d | |
begin | |
literal_listener_example = x -> literal_listener(x, example_threshold, height_density, height_cumulative) | |
plot(0:250, literal_listener_example, | |
legend = false, title ="Updated belief of literal listener", | |
xlabel = "height", ylabel = "probability") | |
end | |
# ╔═╡ 00b7cd46-cda2-11ea-1938-99cb36b62d8f | |
function expected_success(threshold, scale_points, densityf, cumulativef) | |
priors = densityf.(scale_points) | |
beliefs = map(scale_points) do x | |
if x < threshold | |
densityf(x) | |
else | |
literal_listener(x, threshold, densityf, cumulativef) | |
end | |
end | |
sum(priors .* beliefs) | |
end | |
# ╔═╡ 88e604d4-cda3-11ea-3fd5-c7fe2cd3bb13 | |
scale_points = 0:250 | |
# ╔═╡ bd001390-cda3-11ea-2b30-6feee568e1e5 | |
begin | |
es = expected_success(example_threshold, scale_points, height_density, height_cumulative) | |
md""" | |
Expected success for example threshold: $(round(es, digits=5)) | |
""" | |
end | |
# ╔═╡ b0dc3344-cda5-11ea-2ab0-696c1a733d5f | |
function utility(threshold, scale_points, coverage_parameter, densityf, cumulativef) | |
ES = expected_success(threshold, scale_points, densityf, cumulativef) | |
cumulative = coverage_parameter * cumulativef(threshold) | |
ES + cumulative | |
end | |
# ╔═╡ a9bdf5c4-cdab-11ea-1849-c3e86339b76a | |
coverage = 0 | |
# ╔═╡ 938db6e6-cdaa-11ea-0135-f97ab57b30e1 | |
function probability_threshold(threshold, scale_points, lambda, coverage_parameter, densityf, cumulativef) | |
function threshold_value(t) | |
ut = utility(t, scale_points, coverage_parameter, densityf, cumulativef) | |
exp(lambda * ut) | |
end | |
numerator = threshold_value(threshold) | |
denominator = sum(threshold_value.(scale_points)) | |
numerator / denominator | |
end | |
# ╔═╡ a711a104-cdab-11ea-2105-d52d829b81e5 | |
lambda = 50 | |
# ╔═╡ 146fb10a-cdac-11ea-0f6d-3d4a2783f208 | |
threshold_probabilities = map(scale_points) do x | |
probability_threshold(x, scale_points, lambda, coverage, height_density, height_cumulative) | |
end | |
# ╔═╡ ad20f22e-cdac-11ea-0d1a-8f76e6e0e36d | |
plot(scale_points, threshold_probabilities, | |
legend = false, xlabel = "threshold", ylabel = "probability of treshold", | |
title = "Probability that height x is used as the threshold", | |
ylim=(0.0, first(findmax(threshold_probabilities)) + 0.001)) | |
# ╔═╡ 23750612-cdab-11ea-2d1c-2b9a9bd02c28 | |
function use_adjective(degree, scale_point, threshold_probabilities) | |
probability = map(1:length(scale_points)) do i | |
x = scale_points[i] | |
if x <= degree | |
threshold_probabilities[i] | |
else | |
0 | |
end | |
end | |
sum(probability) | |
end | |
# ╔═╡ 7633f6f0-cdaf-11ea-3511-0566c8ac0824 | |
plot(scale_points, x -> use_adjective(x, scale_points, threshold_probabilities), | |
legend = false, xlabel = "height", ylabel = "probability", | |
title = "Probability that a person of height X is described as tall") | |
# ╔═╡ Cell order: | |
# ╟─e6fab282-cd9d-11ea-2efb-156e0cee2c25 | |
# ╟─277190c8-cd9d-11ea-0aee-7952ef163cb2 | |
# ╟─7d1c8156-cd9d-11ea-1e94-816f1ab3173d | |
# ╟─9bd38234-cd9d-11ea-1747-09a10ca39421 | |
# ╟─a64fe396-cd9e-11ea-3a4f-97ef4f9b29b3 | |
# ╟─be288928-cd9e-11ea-23fb-1500bd9d535f | |
# ╟─47535af4-cd9f-11ea-0542-8b04b08f36a7 | |
# ╟─d48144f8-cd9e-11ea-0daf-81c5bda3ba7c | |
# ╠═02c0fd3c-cd9e-11ea-09ac-351343aa0c98 | |
# ╟─a642e89c-cda0-11ea-2466-a7f173352f99 | |
# ╟─82e11130-cda0-11ea-3099-db1268bdf72d | |
# ╟─df623e2e-cda1-11ea-3fea-6d9947eda4c7 | |
# ╠═00b7cd46-cda2-11ea-1938-99cb36b62d8f | |
# ╟─88e604d4-cda3-11ea-3fd5-c7fe2cd3bb13 | |
# ╟─bd001390-cda3-11ea-2b30-6feee568e1e5 | |
# ╠═b0dc3344-cda5-11ea-2ab0-696c1a733d5f | |
# ╠═a9bdf5c4-cdab-11ea-1849-c3e86339b76a | |
# ╠═938db6e6-cdaa-11ea-0135-f97ab57b30e1 | |
# ╠═a711a104-cdab-11ea-2105-d52d829b81e5 | |
# ╟─c3c288a0-cdaf-11ea-2171-edff8c53752d | |
# ╠═146fb10a-cdac-11ea-0f6d-3d4a2783f208 | |
# ╟─ad20f22e-cdac-11ea-0d1a-8f76e6e0e36d | |
# ╠═23750612-cdab-11ea-2d1c-2b9a9bd02c28 | |
# ╟─7633f6f0-cdaf-11ea-3511-0566c8ac0824 | |
# ╟─1e810430-cd9f-11ea-0905-0936032f0267 | |
# ╠═5e827892-cd9d-11ea-1cda-1d8fc9dbd8d2 | |
# ╠═9f2ebd74-cd9d-11ea-2f13-cfe68a07c244 | |
# ╠═04153814-cd9f-11ea-370a-85097498239b |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment