card sort task (case 5345)

How would I create a card sort task in MediaLab? The idea is to present them with a bunch of words and have them sort the words into however many groups they want based on whatever criteria they choose.

For example, if I wanted to know how people think about/categorize holidays, I could give them a bunch of holidays and have them sort into groups they have created. One participant might want to create two groups, one called "holidays to spend with family" and one called "holidays to spend with friends". However, another participant might want to create 3 groups, once called "free holidays", "cheap holidays", and "expensive holidays".

So I'm giving them a set number of words, but they are creating as many categories as they want. It'd be nice to have the words like cards and sort them as such.

Thanks for any advice.

Hi Melissa Abo!

I don’t know of anyone who has done a card sort task in MediaLab, but I hope that I can give you a few leads. In an different thread about creating a version of the Wisconsin Card Sorting Task, Blair Jarvis hinted that an online or Java based version might be adaptable to MediaLab. Therefore, your best bet might be to find an online version of the task. After some Google searching, I found a free, public domain category sorting task program called WebCAT ([url]http://zing.ncsl.nist.gov/WebTools/WebCAT/overview.html)[/url]. You might consider downloading it and seeing if it would fit your purposes. If so, I’d love to continue the discussion about how it might be adapted to work with MediaLab.

Good luck!

Thanks for your help! I will look into it, but it seems like connecting the website and the data to MediaLab might be complicated and/or not possible. I may have to program this study in Qualtrics instead. I've not used it before, but other people in our lab tell me that it can do a card sort task.

Melissa,

Have you ever used a[I] responses.xls[/I] file before? One possible approach would be to use scale response items where the labels would be determined dynamically during the session (e.g., "holidays to spend with friends", "holidays to spend with family"). You could use initial fill in the blank items to capture the participants' "categories". Using the responses.xls file, you could skip them to the sorting task if they enter "done". After that, you could assign those fill-in-the-blank responses as labels for the scale responses. You could also use multiple response items if you wanted subjects to be able to check more than one category. Just a thought.

I'm sorry; I'm afraid I'm not understanding, perhaps because I"ve not used responses.xls. I see how they could create category names with fill-in-the-blank type questions, but I don't understand how they would then sort the words into those categories using a 'scale' type question?

I think when they are sorting the words into categories, they really need to have all the words and all the categories on the same screen. That way they wouldn't have to remember something like "did I already list 'St Patricks Day' as a cheap or expensive holiday?"

Thanks!

If I understand Blair, you will want to create a response file like the example attached to this post. You will also need to create a MediaLab Questionnaire file; I will have to describe it for now because I am away from my computer with MediaLab, but I will post an example later this evening.

Imagine your questionnaire file has 11 items in it: 10 fill in the blank questions and called “fb1” through “fb10” and a multiple response item called “sort.” Each of the fill in the blank questions would be worded like this:

“For the holiday St. Patrick’s Day, please type a phrase that you would use to describe it in the space below. Please continue to type a new phrase in each new space as long as you can think of phrases. When you can’t think of any more phrases to describe St. Partick’s Day, type the word done in all lower-case letters in the space.”

Let’s assume that no participant will list more than 10 phrases, meaning we won’t run out of questions for any participant. Each phrase will be recorded in the row of the attached response.xls file with the corresponding variable name. It also has skip instructions that should move the participant to the next fill in the blank item as long as the participant types anything other than done. If they type done, the participant gets skipped to the “sort” multiple response item in the questionnaire file.

For the “sort” item, click on the Details option. Once there enter <fbx> in each of the “text label” fields of the first 10 rows; replace the x with numbers 1-10. This should show the participant whatever responses he or she entered for the fill in the blank questions.

I hope this helps clarify Blair’s answer.

Cheers.

Psychboiler

Here is the example questionnaire file I promised.

Please let me know if this and my previous post makes sense and if this is closer to what you were looking for.

Cheers!

Hi Psychboiler,

Thank you so much for your efforts to help me. I'm afraid what I'm trying to do is a bit different. Instead of finding the best way to describe something (like St Patrick's Day), we really just want Ps to put holidays into a sort of a pile. For example, holidays would be listed on the left side of the computer screen, and empty boxes would be on the right side of the computer screen. Ps could click and drag holidays from the left into boxes on the right and label the boxes. I'll paste our instructions to Ps below...

Next, you will be asked to sort a list of drugs into groups. First, we would like you to complete an example sorting task. In this example sorting task, you will have a list of 15 different holidays commonly celebrated in America. Your task is to sort these holidays into groups based upon your thoughts and feelings about these holidays. You are to place holidays about which you have similar thoughts and feelings into the same group and holidays about which you have different thoughts and feelings should be sorted into different groups.

Press the space bar to continue

[new screen]

In order to complete this task, you must first decide how many groups you want to have. Remember, you may use as many or as few groups as you desire, but each holiday must be placed in a group.

Father's Day
Veterans Day
Labor Day
Independence Day
New Year's Day
April Fool's Day
Easter
Christmas Day
Columbus Day
Martin Luther King Day
Thanksgiving Day
President's Day
Memorial Day
St. Patrick's Day
Mardi Gras

Press the space bar to continue

[new screen]

For example, given the list of holidays, I came up with the following groups:

[LIST][*]holidays to spend with family: Thanksgiving, Christmas, Easter, Father's day; [*]holidays to spend with friends: New Years, Memorial day, Labor day, Mardi Gras, St. Patrick's Day, Independence Day; [*]quiet holidays: MLK Day, Presidents day, columbus day, veteran's day, st, patricks' day[/LIST]

Ideally, Ps would first decide how many groups they want. Then on the next page, the holidays would be on the left, and the number of boxes on the right would correspond to how many groups they wanted to have. I can do this in Qualtrics using their "rank, group, and order" type question. However, Qualtrics doesn't have RT questions. So, I'd prefer to use MediaLab/DirectRT if it's possible to program this categorization task.

Thanks again for your help,

hi Melissa.
Interesting study you're planning. It seems that you would like to present a fixed list of holidays, such as the one you posted earlier, rather than allowing each participant to generate the list of holidays themselves.

At present, to my knowledge MediaLab and DirectRT do not enable participants to drag and drop an object from one side of the screen to the next, such as each holiday manually grouped into meaningful clusters.

However, it is possible to show the full list, provide fixed scale or open-ended response box and then instruct each to "enter a 1 next to all the holidays during you spend with family" and "enter a 2 next to all of the holidays during which you spend with friends" etc.?

The above is just one example; it assumes that you prescribe the sort groups (family, friends, quiet, etc)

If you want to leave the number of sort groups as flexible as possible (are not steering their responses into family,friend, etc. groups, you could allow a given participant to create up to 15 groups (because it would be possible to assign each holiday to its own group), you could do that, too.

You could also instruct them to limit their groups to at least 2 but no more than say 8.

Either way, the approach is possible, would keep all holidays listed on one page, would provide sorting flexibility, and would record how long it took a participant to complete the task.

Is something like the above tenable for your research?

Thanks so much Timdritchie for the good idea! Unfortunately, since I'll be working with undergrads, I'm not sure that they will be really good at doing that without errors. If they are more sort of physically sorting the words, I think it's easier for them to see which words they've put in which groups. In other words, I have no faith in my Ps, lol.

Hi Melissa,

Thanks for the kind words. It's too bad that I we couldn't come up with something created directly in MediaLab or DirectRT to fit exactly what you needed. Even though it has been a while since my last post, I wanted to share an external application that may be of interest to you.

I recently found out about a card sort task with an online interface. It's called MOIST (Modifiable Online Interface for Sorting Tasks; you know how academics love acronyms that turn into actual words!). It was created by Simon Blanchard for his dissertation. The cool thing about it is that participants can sort items into whatever groups they want, even using the same item in multiple group, and later name the group themselves. This sounds very close to what you wanted.

Here is the URL for his project page:

[url]http://www9.georgetown.edu/faculty/sjb247/[/url]

Check it out!

Psychboiler

Thanks so much Psychboiler! That's perfect!!! I will try to set it up so that they will do the card sort first and then when they exit out of it, I will have already set up the MediaLab/DirectRT portion for them to do the RT tasks. Merging the data shouldn't take me too long. This is great!

Melissa,

That is the first squee I got since posting on the boards here, so I will take it that you like my suggestion. ;) I'm glad this looks like what you were looking for and that our time on this forum thread proved useful. Let me know how this program works out for you and if anything else comes up as you put your study together.

Cheers!

Psychboiler

Thanks Psychboiler :) It's working out fine so far since Ps will be doing the MOIST test first and then DirectRT. However, if that had been opposite, I'd have run into an ID issue since MOIST generates it's own user ID. So what I'm doing is having Ps do MOIST and at the end when it tells them their user ID, I will manually type that in when I set them up on MediaLab/DirectRT so that I'll know which MOIST data goes with which MediaLab/DirectRT data. Would be cool if you guys could work with the MOIST folks sometime to do some sort of ID sharing :)

Hi Melissa,

Thanks for the update. I'm glad to see that you not only got to try out MOIST, but that you are able to use it in concert with MediaLab to conduct your study. I agree that it would be neat to pursue trying to integrate the two programs more seamlessly. Were you thinking of something beyond sharing a common participant ID? I ask because I am not as familiar with what MOIST can do, so it would be great if you could share some thoughts about how to get the most out of MOIST and MediaLab.

Cheers.

Psychboiler

Sorry for the delayed response, we've been working on the MOIST thing. MOIST creates its own particpant ID at the end of the task. This makes using it more complicated :( We ended up deciding not to use it because it just didn't work out. We would have either had to keep track of two different IDs per participant or we'd have had to use MOIST first and then manually put in the ID it generated into MediaLab when they had finished the MOIST part. That task order might work out for other studies, but we really need to start with MediaLab or Qualtrics for this one.

Also, the data goes to the creator and manager of MOIST, and he then emails it to the researchers. He's very good at emailing back quickly, but I think if we were going to start using MOIST and MediaLab together on a frequent basis, it'd be much more convenient if we could access the data ourselves.

So, I think we are going to end up just not doing the card sort task. It's too bad cause I think the task has a lot of potential, and I really like the user-friendliness of MOIST.

Thank you for your help though!

Hi Melissa,

Sorry to hear that the MOIST software seems to cause more hassle than it's worth for your project. Do you still plan on using MediaLab as part of your project? We'd love to hear about how you used MediaLab or provide any further help that you may need.

Don't be a stranger!

Jason (formerly Psychboiler)
Empirisoft Software Support Specialist

[QUOTE=Melissa Abo;19617]Sorry for the delayed response, we've been working on the MOIST thing. MOIST creates its own particpant ID at the end of the task. This makes using it more complicated :( We ended up deciding not to use it because it just didn't work out. We would have either had to keep track of two different IDs per participant or we'd have had to use MOIST first and then manually put in the ID it generated into MediaLab when they had finished the MOIST part. That task order might work out for other studies, but we really need to start with MediaLab or Qualtrics for this one.

Also, the data goes to the creator and manager of MOIST, and he then emails it to the researchers. He's very good at emailing back quickly, but I think if we were going to start using MOIST and MediaLab together on a frequent basis, it'd be much more convenient if we could access the data ourselves.

So, I think we are going to end up just not doing the card sort task. It's too bad cause I think the task has a lot of potential, and I really like the user-friendliness of MOIST.

Thank you for your help though![/QUOTE]

Oh, the wonders of the internet. I am Simon Blanchard, who administers MOIST (now called Cardsorting.net) and found this internet thread. Melissa was one of the first to find me (thank you Jason for the recommendation), and it's gone a long way since then.

My developers and I have actually addressed the features that you requested.
[LIST=1][*]You can create the tasks on your own (kind of like in Qualtrics)[*]You can download the data on your own[*]You can send IDs to the sorting task (so you can link with a survey administered before it)[*]You can send IDs to a survey from the sorting task (so you can link with a survey)[/LIST]

It took me a while to get to this, but I'd be happy to help you out or any other academics who's in need for a sorting task. Feel free to contact me, my info is at [URL="http://www.cardsorting.net"]www.cardsorting.net[/URL].

Best,

Simon

Simon,

That's awesome! Thank you for the update.