RSS

Ex­pected value

TagLast edit: 14 Jul 2022 15:20 UTC by Pablo

The expected value of an act is the sum of the value of each of its possible outcomes multiplied by their probability of occurring.

Illustration

Consider a person choosing between two medical interventions. The first intervention is a drug that succeeds in 60% of cases, and that gives an extra year of healthy life when it succeeds, and has no impact if it fails.

In the case of this drug, there are only two outcomes: success and failure. So the expected value is:

(1 year of life × 60%) + (0 years of life ×40%) = 0.6 expected years of life

Consider another drug that succeeds with a 40% probability, and gives two years of healthy life when it succeeds, but causes harm equal to half a year of healthy life lost when it fails.

Then the expected benefit of this intervention is:

(2 years of life × 40%)−(0.5 years of life × 60%) = 0.5 expected years of life

Over many cases, the first drug will likely provide more years of healthy life than the second. So if they cost the same, funding the first drug would add more healthy years of life on average.

Further reading

Conley, Sean (2016) Deworming might have huge impact, but might have close to zero impact, The GiveWell Blog, July 26.
An example of research using expected value thinking.

Karnofsky, Holden (2011) Why we can’t take expected value estimates literally (even when they’re unbiased), The GiveWell Blog, August 18 (updated 25 July 2016).
A caution about taking applied expected value estimates literally.

Todd, Benjamin (2021) Expected value: how can we make a difference when we’re uncertain what’s true?, 80,000 Hours, September.
An introduction to the concept of expected value.

Tomasik, Brian (2006) Does vegetarianism make a difference?, Essays on Reducing Suffering (updated 25 January 2014).
Another example of expected value reasoning.

Wikipedia (2010) Von Neumann–Morgenstern utility theorem, Wikipedia, April 10 (updated 13 February 2021‎).
For proofs that rational agents should select projects with the highest expected value (note that this does not imply economic risk neutrality).

Related entries

alternatives to expected value theory | altruistic wager | cluelessness | credence | decision theory | moral uncertainty | risk aversion | value of information

Causes and Uncer­tainty: Re­think­ing Value in Expectation

Bob Fischer11 Oct 2023 9:15 UTC
214 points
29 comments15 min readEA link

How bad would hu­man ex­tinc­tion be?

arvomm23 Oct 2023 12:01 UTC
131 points
25 comments18 min readEA link

Ex­pected value the­ory is fa­nat­i­cal, but that’s a good thing

HaydenW21 Sep 2020 8:48 UTC
57 points
20 comments5 min readEA link

The Fu­ture Might Not Be So Great

Jacy30 Jun 2022 13:01 UTC
140 points
118 comments32 min readEA link
(www.sentienceinstitute.org)

Ex­pected im­pact of a ca­reer in AI safety un­der differ­ent opinions

Jordan Taylor14 Jun 2022 14:25 UTC
42 points
16 comments11 min readEA link

Five steps for quan­tify­ing spec­u­la­tive interventions

NunoSempere18 Feb 2022 20:39 UTC
91 points
8 comments11 min readEA link

Po­ten­tial down­sides of us­ing ex­plicit probabilities

MichaelA20 Jan 2020 2:14 UTC
57 points
22 comments18 min readEA link

The Max­i­mum Im­pact Fund is now the Top Char­i­ties Fund

GiveWell8 Sep 2022 19:31 UTC
57 points
3 comments1 min readEA link
(blog.givewell.org)

Thought ex­per­i­ment: Trad­ing off risk, in­tra­gen­er­a­tional and in­ter­gen­er­a­tional in­equal­ity, and fairness

Jobst Heitzig (vodle.it)2 Sep 2023 23:32 UTC
9 points
6 comments3 min readEA link

1/​E(X) is not E(1/​X)

EdoArad9 Nov 2023 10:22 UTC
55 points
14 comments1 min readEA link

Ar­gu­ments for util­i­tar­i­anism are im­pos­si­bil­ity ar­gu­ments un­der un­bounded prospects

MichaelStJules7 Oct 2023 21:09 UTC
39 points
48 comments1 min readEA link

In Cost-Effec­tive­ness Analy­ses, when should we take the Ex­pected Value?

EdoArad13 Mar 2024 14:49 UTC
15 points
4 comments2 min readEA link

[Question] What is the marginal im­pact of a small dona­tion to an EA Fund?

BrownHairedEevee23 Nov 2020 7:09 UTC
12 points
4 comments1 min readEA link

[Question] Pro­tect­ing EV calcu­la­tions against mo­ti­vated reasoning

freedomandutility12 Dec 2022 12:29 UTC
14 points
0 comments1 min readEA link

Ex­pected value and un­cer­tainty with­out full Monte Carlo simulations

Vasco Grilo5 Jan 2024 8:57 UTC
12 points
2 comments2 min readEA link

High-risk high-re­ward philan­thropy: ap­ply­ing ven­ture cap­i­tal con­cepts to do­ing good

James Özden4 Jul 2022 13:00 UTC
15 points
2 comments9 min readEA link

[Question] Mun­dane trou­ble with EV /​ utility

Ben Esche3 Apr 2021 7:51 UTC
37 points
17 comments2 min readEA link

A Para­dox for Tiny Prob­a­bil­ities and Enor­mous Values

Global Priorities Institute1 Jul 2021 7:00 UTC
5 points
1 comment1 min readEA link
(globalprioritiesinstitute.org)

Se­quence think­ing vs. cluster thinking

GiveWell25 Jul 2016 10:43 UTC
17 points
0 comments28 min readEA link
(blog.givewell.org)

[Question] What rea­son is there NOT to ac­cept Pas­cal’s Wager?

Transient Altruist4 Aug 2022 14:29 UTC
30 points
82 comments1 min readEA link

Shap­ley val­ues: an in­tro­duc­tory example

Stan Pinsent12 Nov 2023 13:35 UTC
9 points
0 comments3 min readEA link

Prob­a­bil­ities, Pri­ori­ti­za­tion, and ‘Bayesian Mind­set’

Violet Hour4 Apr 2023 10:16 UTC
55 points
6 comments24 min readEA link

Sum­mary: In Defence of Fa­nat­i­cism (Hay­den Wilk­in­son)

Nicholas Kruus15 Jan 2024 14:21 UTC
26 points
2 comments6 min readEA link

Cos’è il val­ore at­teso (ex­pected value)

EA Italy31 Dec 2022 4:21 UTC
1 point
0 comments1 min readEA link

In ex­tremely high-stakes sce­nar­ios, it’s ok not to max­imise ex­pected utility

Pivocajs13 Oct 2022 1:13 UTC
14 points
11 comments1 min readEA link

Ex­pected Value

james31 Jul 2020 13:59 UTC
22 points
2 comments1 min readEA link

SBF, ex­treme risk-tak­ing, ex­pected value, and effec­tive altruism

vipulnaik13 Nov 2022 17:44 UTC
73 points
17 comments26 min readEA link

Ex­pected Value

EA Handbook23 Jul 2022 23:00 UTC
8 points
3 comments1 min readEA link
(non-trivial.org)

Bit­ing the Bul­let on Pas­cal’s Mugging

Matthew Barber28 Aug 2022 15:30 UTC
5 points
0 comments3 min readEA link

Op­tion Value, an In­tro­duc­tory Guide

Caleb_Maresca21 Feb 2020 14:45 UTC
31 points
3 comments7 min readEA link

Con­cerns/​Thoughts over in­ter­na­tional aid, longter­mism and philo­soph­i­cal notes on speak­ing with Larry Temkin.

Ben Yeoh27 Jul 2022 19:51 UTC
35 points
1 comment9 min readEA link

My notes on: Why we can’t take ex­pected value es­ti­mates liter­ally (even when they’re un­bi­ased)

Vasco Grilo18 Apr 2022 13:10 UTC
14 points
0 comments3 min readEA link

Do Brains Con­tain Many Con­scious Sub­sys­tems? If So, Should We Act Differ­ently?

Bob Fischer5 Dec 2022 11:56 UTC
90 points
8 comments21 min readEA link

FTX/​CEA—show us your num­bers!

Jack Lewars18 Apr 2022 12:05 UTC
159 points
109 comments2 min readEA link

A Gen­eral Treat­ment of the Mo­ral Value of Information

SamNolan17 Jul 2021 22:50 UTC
16 points
0 comments9 min readEA link

[Question] What should I ask Alan Há­jek, philoso­pher of prob­a­bil­ity, Bayesi­anism, ex­pected value and coun­ter­fat­u­als?

Robert_Wiblin1 Jul 2022 13:23 UTC
25 points
12 comments1 min readEA link

Con­se­quen­tial­ism and Cluelessness

Richard Y Chappell17 Oct 2022 18:57 UTC
32 points
5 comments9 min readEA link
(rychappell.substack.com)

EV Max­i­miza­tion for Humans

Sharmake3 Sep 2022 23:44 UTC
12 points
0 comments4 min readEA link

How to de­ter­mine dis­tri­bu­tion pa­ram­e­ters from quantiles

Vasco Grilo30 May 2022 15:20 UTC
23 points
6 comments1 min readEA link

Should strong longter­mists re­ally want to min­i­mize ex­is­ten­tial risk?

tobycrisford4 Dec 2022 16:56 UTC
31 points
9 comments4 min readEA link

Fa­nat­i­cal EAs should sup­port very weird projects

Derek Shiller30 Jun 2022 12:07 UTC
66 points
42 comments9 min readEA link

A model about the effect of to­tal ex­is­ten­tial risk on ca­reer choice

Jonas Moss10 Sep 2022 7:18 UTC
12 points
4 comments2 min readEA link

[Question] When should the in­verse-var­i­ance method be ap­plied to dis­tri­bu­tions?

Vasco Grilo14 Jun 2022 14:33 UTC
5 points
6 comments1 min readEA link

What is so wrong with the “dog­matic” solu­tion to reck­less­ness?

tobycrisford11 Feb 2023 18:29 UTC
25 points
31 comments7 min readEA link

We can’t put num­bers on ev­ery­thing and try­ing to weak­ens our col­lec­tive epistemics

ConcernedEAs8 Mar 2023 15:09 UTC
9 points
0 comments11 min readEA link