In this assignment, you will have to implement a modification of a learning by feedback task. In this task, participants are repeatedly presented with 2 options, one on the left and one on the right side of the screen, and have to choose between one of them (by either pressing ‘q’ or ‘p’ on the keyboard).
After choosing, they receive a numerical feedback (lasting 2 seconds) corresponding to each option, but they only ‘collect’ the reward corresponding to the option that they chose. At the end of the experiment, they receive a monetary bonus corresponding to the accumulated rewards. In this way, they are incentivised in learning and choosing the options with higher rewards.
In total, there are 4 different options in a block (A, B, C, and D), identified with some abstract figure. You can find the figures that identify the options in the dropbox folder.
Here, we are going to reduce the number of trials and blocks, compared to the original experiment.
In a block (made of 40 trials), each option is presented for a total of 20 times. These are the feedback values associated with the four options A, B, C, and D:
A:
round(rnorm(20, mean=36, sd=5))## [1] 38 36 33 33 36 35 35 33 40 33 32 44 37 40 43 40 34 25 37 37B:
round(rnorm(20, mean=40, sd=5))## [1] 39 31 33 43 40 44 42 43 37 40 36 44 40 33 38 41 36 44 43 44C:
round(rnorm(20, mean=50, sd=5))## [1] 49 47 48 44 52 40 53 58 50 44 51 47 45 50 51 48 59 48 51 47D:
round(rnorm(20, mean=54, sd=5))## [1] 47 48 50 51 56 54 61 50 56 54 49 51 51 59 61 47 48 56 54 59The mean rewards of the options in each block were 36, 40, 50, and 54 for options A, B, C, and D, respectively. And the order of the feedback was randomised for each option.
However, in a specific trial, only 2 of them are presented at a time: only four of the six possible pairs of options were presented: AB, AC, BD, and CD (but not AD and BC).
Even though there were 3 blocks in the original experiment, we are only going to implement 1 test block (and at the end we are going to add a practice block).
Create a new Inquisit script file and name it
wpa_7_FirstLast.iqx (with your First and Last name). This
is the file you will email to me at the end of the class.
Before starting, change the following default settings:
/screencolor = (238, 238, 238),
/txbgcolor = (238, 238, 238). This will make sure the
background will match the stimuli background.
Exercise 1a: Create four picture
elements, called A, B, C,
D. They should correspond to the .png figures in the
material folder, that you should download on your computer first (be
sure to specify the file path correctly). For now, A and D should be
presented on the left, and B and C on the right of the screen (set the
left/right position at 35%/65%). Set the vertical position to 50% for
all options. Set both height and width to 40% for all stimuli.
Preferably you should use the parameters element to set these general
stimuli settings.
Exercise 1b: Create four list elements,
called rewardA, rewardB, rewardC,
and rewardD. They should contain the reward feedback listed
above. Note that in this case we can use the list and not the item
element because we can define a list of numbers that we can sample from.
Have a look at the Examples in the list element Help page in Inquisit to
see how to write the rewards in the /items attribute. Set
the /resetinterval attribute to 1, which means that when we
are sampling with no replacement, there will be a reset of the samples
at the beginning of each block.
Exercise 1c: Create four text elements,
called feedbackA, feedbackB,
feedbackC, and feedbackD. They should have as
items the corresponding item elements. The selection method should be
without replacement. The position should be the same of the
corresponding pictures. Also, you should set the fontsize to 10%. Now,
we have to explicitely extract a number from the reward lists, for
example: /items = ("<%list.rewardA.nextvalue%>") for
feedbackA.
Exercise 2a: Define 4 trial elements,
called choice_AB, choice_AC,
choice_BD, and choice_CD. They should present,
at the begininning of the trial, the two corresponding picture elements.
Set the valid and correct response in each trial (remember that the
correct option is the one with highest mean value and that the mean
values are 36, 40, 50, and 54 for options A, B, C, and D, respectively).
Finally, include a response deadline at 3 seconds and a pre-trial pause
of 500 msec.
Exercise 2b: Define 4 trial elements
called feedback_AB, feedback_AC,
feedback_BD, and feedback_CD. They should
present the two corresponding feedback text elements. Feedback should be
presented for 2 seconds, right after a response is made (the pretrial
pause should be 0 for the feedback trials).
Exercise 2c: In order to make the feedback appear
right after a response is made, use the responsetrial
argument in the choice_AB, choice_AC,
choice_BD, and choice_CD elements. This
argument takes 2 values: the first is the response and the second is the
trial element. So you will have to set it twice (for both possible
responses “q” and “p”). Note that this is similar to using the
/branch attribute, but we can specify it already
conditionally to a specific response (instead of conditions such as “the
response latency was higher than X seconds”).
Exercise 3a: Define a block element
called choice_block made of 40 trials, where
choice_AB, choice_AC, choice_BD,
and choice_CD are presented without replacement.
Exercise 3b: Define an expt element
made of 1 block (the choice_block).
Congratulations, you should now have a working Experiment. Try it out before you attempt the rest of the Exercises.
Now, we want to add a practice_block at the beginning,
and allow participants to continue with the choice_block
only if they reached a minimum of 60% accuracy. Also, we want to add
instructions to warn the participants that they are going to start
either a practice or a test block, and that the rewards will only be
accumalted in the test block.
You can decide what happens after participants fail to reach 60% accuracy: they can either have the chance to repeat the practice block or the experiment could stop here.
Exercise 4a: Implement a practice_block
made of 20 trials, and let the participants repeat it until they reach
60% accuracy.
Exercise 4b (bonus): Set the limit of
practice_block repetitions to 2, and otherwise let the
experiment stop, preferably with a final message that tells the
participant that the experiment has ended because they did not reach the
minimum accuracy required.
Now, the options are always shown at fixed positions on the screen. In a real experiment, however, we most likely would want to randomise this.
<list left_or_right>
/items = (1, 2)
/select = random
/replace = true
</list>Here I created a list that can be either 1 or 2.
Exercise 5: Sample from it (using
list.left_or_right.nextvalue) at the beginning of each
trial (using \ontrialbegin) and change the picture’s
position (by setting, for eaxample
picture.A.hposition = parameters.x_left; picture.B.hposition = parameters.x_right
when the list value is 1 and otherwise
picture.B.hposition = parameters.x_left; picture.A.hposition = parameters.x_right).
Note that you will have to move the feedback as well (using the same
method).
We also want to show to the participants their final reward.
# only inaccurate responses
sum(round(rnorm(20, mean=36, sd=5))) + sum(round(rnorm(10, mean=40, sd=5))) + sum(round(rnorm(10, mean=54, sd=5)))## [1] 1672# only accurate responses
sum(round(rnorm(10, mean=40, sd=5))) + sum(round(rnorm(10, mean=50, sd=5))) + sum(round(rnorm(20, mean=54, sd=5)))## [1] 1973Here I simulated the reward amount of 2 subjects, one that would only choose the lowest paying option, and one that would only choose the highest paying one.
At the end of the experiment, we want to give participants some bonus based on their responses, and therefore scale these accumulated rewards to something we could give them in CHF. For example, we could divide by 1000 and then round to 10th of rappen, so for example:
final_reward = sum(round(rnorm(20, mean=36, sd=5))) + sum(round(rnorm(10, mean=40, sd=5))) + sum(round(rnorm(10, mean=54, sd=5)))
final_reward## [1] 1668round(final_reward/100)/10## [1] 1.7In order to do that, we can use the expression element. This can go together witht the values and parameters elements at the beginning of the script, but can be used to specify an operation that we want to be made when we call it:
<expressions>
/transform_reward = round(values.accumulated_reward/100)/10
</expressions>Exercise 6a: Create a
accumulated_reward value that we update at the end of each
trial, based on participants’ response.
Exercise 6b: Create a new variable called
values.accumulated_reward_CHF which is reset at the end of
each trial using the transform_reward expression.
Exercise 6c: Add a message at the end of the experiment that tells participants how much money they have won.
Until now, we didn’t really care too much about how the data would
look like in the .iqdat output file.
However, we might need to modify the columns to have a better control/understanding of the variables in the experiment.
For example, here we might want to be sure that we are indeed calculating the accuracy in the practice block correctly.
We can start from adding this at the end of your script:
<data>
/columns=[
build, computer.platform, date, time, subject, blocknum, blockcode, trialcode, latency, response, correct, stimulusitem, stimulusitem, list.left_or_right.currentvalue]
/separatefiles = true
/labels = true
</data>Exercise 7: Add columns to track the accumulated rewards (check that they indeed reset at the beginning of the choice block), and the variables you used to keep track of the accuracy in the practice block (check that indeed the choice block only starts when the accuracy is above .6). You might need to run the experimebt a few times in order to check this (and it might be useful for you to learn which options are the correct one: my strategy is to avoid the pink stimulus when it’s presented and to choose the violet one when it’s presented. Note that they are never presented in the same trial so this is a sufficient strategy).
You can send your assignment to my email address: laura.fontanesi@unibas.ch.
For this assignment, please send me the Inquisit script. If you want
to add comments, you can do that by adding the hashtag # at
the beginning of the line.