Dan Weijers
Victoria University of Wellington
July 2011
The Experience Machine is Dead, Long
Live the Experience Machine!
Q: Does Nozick’s experience machine thought experiment (EMTE) elicit biased responses?
Q: Should Nozick’s EMTE be used as evidence in arguments about wellbeing?
Q: Should any version of the EMTE be used as evidence in arguments about wellbeing?
Q: What do EMTEs tell us about the value of pleasure and the plausibility of hedonism?
Q: What should EMTEs be used for?
‘Q’s to be ‘A’ed
“Suppose that there were an experience machine that would give you any experience you desired. Superduper neuropsychologists could stimulate your brain so that you would think and feel you were writing a great novel, or making a friend, or reading an interesting book. All the time, you would be floating in a tank, with electrodes attached to your brain. Should you plug into this machine for life…?” Nozick, R. (1974). Anarchy, State, and Utopia, pp 42-45 Smart, J. & Williams, B. (1973). Utilitarianism for and Against , pp 18-21.
Disclaimers:Ignore responsibilities to family etc. Don’t realise it’s not realStill exercise autonomyMachine works perfectly
The Experience Machine
Internalist mental state theories of wellbeing = ‘hedonism’
“Now if pleasure were our greatest good, then we would all volunteer to be hooked for life to this machine… But surely very few people would volunteer.” Kymlicka, W. (1990). Contemporary Political Philosophy: An Introduction, p 13.
1. If pleasure = g. good, then we’d plug in2. We would not plug in3. Therefore, pleasure ≠ g. good
But is premise 1 true? It assumes our judgements about the EMTE are based on
a rational comparison of lots of pleasure with average pleasure
Is premise 2 true for other EMTEs?
The Experience Machine vs. Hedonism
Judgments are affected by intuitions, but it’s hard to say when and by how much because intuitive cognition is opaque How easily does ‘the reason’ come?
It’s also hard to know what causes the intuition (and therefore how useful it is) Reconstruction via reverse engineering Reconstruction often fails when we only consider the factors that
are stipulated in the thought experiment as the possible causes of the intuition.
Since intuitive cognition uses pattern recognition, irrelevant aspects of the triggered past experiences can affect the intuition.
Experiments have caught people out confabulating / being dumbfounded
Biases – The problem with intuitive cognition (and judgments heavily influenced by it)
Intuitions and Judgments
David Sobel: the credibility of intuitions elicited from contemplation of thought experiments can be undermined by “telling a convincing story about the genesis of such intuitions that would explain why we have them while revealing them to be misleading” (2002, p. 244).
Tweaks and reversalsArmchair, hallway, and experimental testing
Debunking Intuitions
Imaginative resistance = consciously, or unconsciously, rejecting any of the stipulations (or implied features) of a thought experiment.Machine underperformanceMachine malfunctionNot worrying about loved ones
Also the opposite - overactive imagination. “floating in a tank, with electrodes attached to
your brain,” “plug into this machine” (My emphasis, Nozick, R. (1974). Anarchy, State, and Utopia, p 42.)
Imaginative Resistance
125 first year philosophy students with Nozick’s scenario: which is the best option to choose and why?
79% (99/125) thought they should not connect 35% (35/99) cited reality, truth, or something related
44% (44/99) of the main justifications given by those who thought they should not connect indicated imaginative resistance bad experiences are required to appreciate good ones or develop
(19%, 19/99) no autonomy or control in the machine, (14%, 14/99) responsibilities to others (4%, 4/99) the machine might break down or not be that good (3%, 3/99)
Other main justifications that were consistent with the thought experiment, but are irrelevant for evaluating the intrinsic value of experiences or reality, included: the machine seems scary or unnatural (8%, 8/99) getting out every two years would be depressing (6%, 6/99)
Test: “Nozick’s” Scenario
93 first year business students using the following survey:It’s 2062 and you are riding a hovertube to town…
1) Ignoring how your family, friends, any other dependents, and society in general might be affected, and assuming that Experience Machines always work perfectly, what is the best thing for you to do for yourself in this situation?Tick only one of these options: O You should accept the spot in an Experience MachineO You should not accept the spot in an Experience Machine2) Briefly explain your choice:
37% (34/93) thought they should connect to an EM
Test: My “Self” Scenario
Far fewer respondents to the Self scenario justified their choice with the reasons: ‘bad experiences are required to appreciate good
experiences or to develop properly’ and ‘you would have no autonomy or control in the machine’
Self vs. Nozick’s Scenario
Self Nozick's0
5
10
15
20
25
30
35
40
'Should Connect to the EM' Re-
sponses
%
Self Nozick's0
5
10
15
20
25
30
35
40
45
50
Irrelevant Main
Judgments
%
16% differencep = 0.004
23% differencep < 0.001
Status quo bias is best defined as an inappropriate preference for things to remain the same Bostrom, N. & Ord, T. (2006). The Reversal Test: Eliminating Status Quo Bias in Applied
Ethics, Ethics, 116: 656-679.
Linked to:Loss aversion (valuing losses more than
equivalent gains in uncertain circumstances) and
The endowment effect (overvaluing what we have and know)
Status Quo
Limitations:Small
samples (24, 80, 24, 24)
Social capital lost in reality (not a mirror of Nozick’s)
De Brigard's (2010) Respondents Staying Connected
Negative Nuetral SQ empahsised
Positive Nuetral0
10
20
30
40
50
60
70
80
90
100
Real Life is…
%
De Brigard, F., 2010. If You Like it, Does it Matter if it’s Real?, Philosophical Psychology, 23(1): 43–57.
Reduce oversensitivity to potential lossesEvolutionary considerations we are loss averse in
conditions of uncertainty Chen M. K., V. Lakshminaryanan, & L. R. Santos, 2006. The Evolution of Our Preferences: Evidence from
Capuchin Monkey Trading Behavior, Journal of Political Economy, 114: 517–537.
Behavioural econ. we value equivalent losses and gains more equally the less we know the target of the decision Bloomfield A. N., J. A. Sager, D. M. Bartels, D. L. Medin, 2006. Caring about Framing Effects, Mind &
Society, 5(2): 123–138.
Therefore, we are more likely to think that the unfamiliar (and therefore risky) experience machine life is going to be a better idea for someone we care less about than ourselves E.g. respondents will be more likely to think that a stranger
should connect to an experience machine than a friend, and a friend more than themselves.
Eliminating Status Quo 1
This hypothesis was tested with 3 new scenarios in which a friend, a cousin, and a stranger are the targetAll of these scenarios were tested on first year business
students at the same time as the Self scenario.Very similar to Self scenario, only now the choice is
about whether the stranger should connect to an EM48% (45/93) decided that connecting to an
experience machine made the life in question better12% more than respondents to the Self scenario.
Test: My “Stranger”, “Cousin” & “Friend” Scenarios
May not be loss aversion that is the causeAre EM judgments becoming more rational as
care factor decreases?
Reducing Loss Aversion by Decreasing the Care Factor
Framing all of the options as equally familiarFraming of reality or the machine as the
status quo makes it the easy, less risky, known, and perhaps default option
A main appeal of the SQ seems heavily linked to our social and relational capitalwe know how to work with the life we have
Eliminating Status Quo 2
Framing of the SQ at a Glance
Noz
ick’
s Se
lf
Friend
Cousin
Stra
nger
DB’s N
eutral
DB’s P
ositi
ve
DB’s N
eutral
SQE
DB's Neg
ativ
e 0
10
20
30
40
50
60
70
80
90
100
Con
nect
%
124 first year philosophy students - Stranger No Status Quo scenario:
A stranger, named Boris, has just found out that he has been regularly switched between a real life and a life of machine-generated experiences (without ever being aware of the switches); 50% of his life has been spent in an Experience Machine and 50% in reality. Nearly all of Boris’ most enjoyable experiences occurred while he was in an Experience Machine and nearly all of his least enjoyable experiences occurred while he was in reality. Boris now has to decide between living the rest of his life in an Experience Machine or in reality (no more switching)…
1) Ignoring how Boris’ family, friends, any other dependents, and society in general might be affected, and assuming that Experience Machines always work perfectly, what is the best thing for Boris to do for himself in this situation?
Tick only one of these options:
O Boris should choose the Experience Machine life
O Boris should choose the real life
Test: My “Stranger NSQ” Scenario
Neutralising the status quo = 12% difference (Stranger NSQ vs. Stranger, p-value = 0.025)
Making the choice on behalf of a stranger = 12% difference (Stranger vs. Self , p-value = 0.035)
Reducing irrelevant factors = 16% difference (Self vs. Nozick’s , p-value = 0.004)
Stranger NSQ vs. Nozick’s = 40% difference (p-value = 0.000)
Reducing Bias in EMTEs
YesIndicated by De Brigard’s resultsIndicated by students’
confabulation/dumbfounding in classShown by my results
Imaginative resistanceStatus Quo Bias
Over-sensitivity to loss/fear of change/inertia
Does Nozick’s (EMTE) elicit biased responses?
NoNozick’s scenario elicits judgements about
40% off the mark due to:Status quo biasImaginative resistanceIneptness (irrelevant main justifications)
The experience machine is dead!
Should Nozick’s EMTE be used in arguments about wellbeing?
But are trained philosophers affected?It might not matterWhose judgments do we want to understand?
Hold on!
Philosophers Students Reasonable people
0
10
20
30
40
50
60
70
Hypothetical Connect % for Stranger NSQ
Philosophers Students Reasonable people
0
10
20
30
40
50
60
70
Hypothetical Connect % for Stranger NSQ 2
Possibly The Stranger NSQ scenario seems like a much
better candidateReality vs. internal experiences (Yes)Makes use of memorable, mysterious and
intriguing concept of the experience machine (Yes)Appears relatively unaffected by bias and other
irrelevant factors (Yes)Not too complicated < half %inept responses of
Nozick’s (OK)Long live the experience machine!
Should any EMTEs be used in arguments about wellbeing?
Not as much as philosophers have thoughtNo agreement on the relative valuesThe EM argument against hedonism needs reassessing
These results do not endorse hedonism22% cited reality as main justification in the Stranger
NSQ scenario If reality has any intrinsic value, then hedonism is false22% citing reality = defeasible evidence that hedonism
is falseThe experience machine should no longer be
considered to provide evidence that hedonism and all other internalist mental state theories of well-being are overwhelmingly likely to be false.
What do EMTEs tell us about pleasure and hedonism?
All of the ETMEs should be used to educate on how to evaluate thought experiments
Long live the experience machine (again)!
Stop the student abuse10 ‘normal’ lecturer spiel on EMTE vs. hedonismExposed connect % = 28%Control connect % = 15%13% difference p-value = 0.032Is this how judgements become widespread
(amongst philosophers)?
What should EMTEs be used for?
How much confabulation is going on in EMTEs?
Do philosophers rationally judge EMTEs?Confirmation bias
How would trained philosophers with no experience of EMTEs respond to my Stranger NSQ scenario?Would experience of Nozick’s EMTE make
much difference?
Interesting Questions
Q: Does Nozick’s experience machine thought experiment (EMTE) elicit biased responses? Yes
Q: Should Nozick’s EMTE be used as evidence in arguments about wellbeing? No
Q: Should any version of the EMTE be used as evidence in arguments about wellbeing? Possibly
Q: What do EMTEs tell us about the value of pleasure and the plausibility of hedonism? Less than thought
Q: What should EMTEs be used for? Education
‘Q’s ‘A’ed
Top Related