10/29: Nicole Ellison

[ehuppert]

Comment below with questions or thoughts about the reading for this week's workshop.

Please make your comments by Wednesday 11:59 PM, and upvote at least five of your peers' comments on Thursday prior to the workshop. You need to use 'thumbs-up' for your reactions to count towards 'top comments,' but you can use other emojis on top of the thumbs up.

[nwrim]

Thanks for presenting in our workshop! I really enjoyed the paper a lot. I think all three categories of "non-click" behavior based on the interview really resonated with me - I am sure that I considered all three when I used to be more active on Facebook.

One thing that I am curious about is how can we scale-up the study of "non-click" behaviors. Contrary to clicking or actively-engaging behavior that leaves digital traces on the platform, I don't think non-click behavior will be easy to track down. I think the usage of eye-tracking was a really clever way to track one's attention and investigate this, but it will be difficult to have thousands of people go through the eye-tracking procedure as they look into their Facebook feeds. Do you have any thoughts on this issue? I think scaling up this can answer a lot of interesting questions, such as "did people reduce their clicking behavior based on 'Audience: the platform' reasons as the algorithms implemented on social network problem got more public attention?"

[mikepackard415]

Thank you for sharing this research! The non-click concept is definitely a bit a of a black hole for researchers using social media data. In much the same way that we might observe what people choose to spend money on or how they choose to spend their time, we don't get to look inside their heads to see what alternatives they considered but didn't select. One thought I'm having is that when advertisers place ads on Facebook, they often pay for a number of impressions, which counts the number of times a browser displayed their ad. I wonder if researchers could somehow develop an analog for the number of times a post is displayed in a browser, and from there could calculate the number of non-clicks a certain post got. There are some immediate issues like data availability and privacy, and these are by no means easy to tackle, but theoretically this seems like it could be one pathway to studying the non-click without relying on eye tracking.

To summarize the question: Do you see an opportunity to study the non-click on a larger scale using some sort of analog to the 'impressions' that are tracked for advertisements?

[JadeBenson]

Thank you for sharing! I thought the "social content" measure and its relationship with clicking was interesting. This measure is calculated by dividing the number of posts by Friends over the total number of posts. I was wondering how you might incorporate social groups into this measure? For example, UChicago has a variety of group pages that generate a lot of content but some of it is anonymized, or from people you're not explicitly "friends with" but share a social group. These posts generate a ton of interaction (and at least dominate my feed). I know this might not apply as much to your study sample (since it was an "older demographic" of non gen-Zers), but I was wondering of how you might incorporate affinity groups and group social interaction into future research?

[bakerwho]

Thanks for sharing your work with us!

I really enjoyed reading your paper - as a computational social scientist who works more with data-ified subjects, it was refreshing to read work that involved interviewing actual people.

In your paper, you make many quantitative observations like - "each percent increase in social content is associated with 1.72% more clicking". Platforms like Facebook to optimize their interfaces and algorithms with pain-staking frequency, and their 'variables of interest' are often along these lines, as they tie in logically with their business models. It struck me how much relatively simpler it is to operationalize optimization problems in terms of click-through rates, ad pricing, and profit - than subtler user preferences like comfort (with a feed item), privacy, inclusivity, or fairness.

I also really appreciated the discussion under 'Design implications of the non-click' - on the plethora of incorrect or even harm-causing ways algorithms could interpret these less obvious user behaviors. With that long piece of context, here's my questions. How should we think about other such non-obvious 'improvements' social algorithms might need to make? How can we balance these improvements with commitments to privacy (i.e. if records of inaction become personal information, that raises new privacy concerns)?

Thank you again!

[mintaow]

Thanks for sharing this interesting work. Similar to Nak Won, I am also very interested in the possibilities to scale up this experiment. In Page 7, you mentioned that you are using Tobii X2-30 to collect eye-tracking data. I did a quick preliminary search and found that prices for eye-tracking devices are surprisingly high and unfriendly to the public or researchers. According to iMotions, the average eye tracker price is around $17,500 and Tobii is among the middle-end range of about $1,000 ~ $ 10,000. (Unfortunately, prices for Tobii devices are given upon inquiries and I haven't verified this number yet. But several consumer reviews also mention the high price).

So my question is, is it possible for us to scale up this experiment using more economic eye-ball tracking solutions such as webcam+SDK, which I guess only costs below $100 per device? Thus, I am really curious about how you decide to use Tobii X2-30 and what are the common methods/devices of including eyeball tracking in user studies. Thanks!!

[sabinahartnett]

Thank you for sharing your research! I am especially interested in media consumption and human-machine interaction and so this added an important nuance to the "dichotomous 'passive versus active' framing" as you put it, of user engagement on Facebook. My question is a bit theoretical but I wonder about the significance of user agency in this study. While the interviews show clear thoughtful/purposeful usage in the time participants were being watched and some participants showed self-awareness of how their participation would effect 'their algorithm'... machines have the ability to elicit addictive-qualities in users which often pulls them into a 'zone' of engagement where the primary goal is continuation rather than engagement (here described in regard to machine gambling but many parallels to social media design).

I'm wondering how much you considered userface design and how the goals of the platform may influence/inform user behavior (especially outside of the study, in more addictive environments where users can scroll unlimited through the timeline)?

[MkramerPsych]

Thank you for sharing this work with us! I have always believed that what we see but don't click says a lot about us, and this is the first paper I've read on the subject. I have three questions:

1) How do you reconcile the "influence market" with your findings? Does the conceptualization of likes as a "currency" complicate the relationship between viewing and clicking? Specifically, how would you go about disentangling "metacommunicative" likes from those driven by self-presentation?

2) It seems like a great deal of the research on Facebook relies heavily on the individual relationship each user has with the service. For some, not receiving a like from a friend can feel incredibly dismissive or outright hurtful, while for others it would have no emotional salience. I am curious - could a measure be developed that categorizes people into those who place a certain amount of weight on their social media profiles and those who don't, and would that measure be useful in this line of research?

3) While I agree with the notion that categorizing non-clicks as passive behavior is intrinsically flawed due to active withholding, I am curious about how the opacity of a person's non-clicks to other Facebook users could be reconciled for future social media research. Most Facebook users don't have the time to see everyone on their friends list who did not like their posts, but it is clear that what we choose not to click contains valuable information. How would you approach finding the social ramifications of the non-click behavior?

[Lynx-jr]

Thanks for presenting in our workshop and bringing this awesome paper! I really enjoyed this mix of eye-tracking techniques and combining interviews to forward our knowledge about how people feel about eye-ing web-contents in general. (I'm not a professional in such fields so it might be a stupid question) From what I heard researchers from MIT and Georgia Tech have been using Mturk to test whether front cameras are able to do what Tobii is doing... Does current smartphones have the capability to catch eye-movement?

[lulululugagaga]

Thanks for your great research. This paper reminds me of a talkshow of Lan Hu. He jokes that when he opens Tik Tok, he would only swipe away those sexy lady videos immediately because he is too shy to watch and would stop at any other pages for a while to decide if he's interested. This might also constitute a reason for non-clicking behaviors. So from my point of view, the reasons of non-clicking behaviors could be really difficult to explain, especially when the dataset is small. It could be just natural neglect, or different level of sensitivity to color. If we can have large experiment and control groups to make comparisons, the result might be interesting.

[skanthan95]

Thank you for presenting at our workshop! I really enjoyed reading your paper. Here are some comments and questions:

• As some of my peers have pointed out, your Facebook feed is largely curated based on the content that you've clicked in the past. For study purposes: If users were presented with their/a Facebook account in a way that wasn't already calibrated based on what they've already clicked in the past, they might click way less often than they would have if they were interacting with their own 'calibrated' account. It could be interesting to see how click and viewing patterns change in this scenario and what influences these changes. I could imagine a situation where an individual might be exposed to content that they actually might like/click, but didn't typically see on their newsfeed due to previous content that they clicked.

• I'm interested in data on clicks in public versus private groups. Unless a Facebook friend is in your private group as well, they wouldn't be able to see that you clicked something on that page. This added layer of privacy might be appealing to some users - knowing that only a subset of people in their friends list could see a record of a particular click. Do you have any data on public vs. private-page clicks?

[LFShan]

Thank you for sharing. I really didn't expect that people spend a similar amount of time on the content they click on and content they don't click. I think one way to upscale the research is to use the amount of time they stay on one screen as a substitute for times the spend looking at certain content. It might be less accurate but maybe the data would be easier to collect overtime and for a large scale.

[a-bosko]

Dr. Ellison,

Thank you for sharing your research! It was very interesting to read about the passive and active uses of social media. It was also very interesting to learn about the differences between clicking and viewing on social media; it is exciting to see research being done on the different ways social media is used and the health outcomes of usage.

My question is, is it possible that there are personality differences between clickers and lurkers? It would be interesting to look at conducting research between personality types and clickers/lurkers. For example, someone who falls closer to extroversion may be more likely to be a clicker, since extroverts are probably more likely to seek attention from others.

Thank you very much!

[PAHADRIANUS]

Thank you for sharing your upfront research results. While I am familiar with a good portion of the survey or experiment methods employed in your study, your incorporation of eye-tracking is astoundingly impressive, and even more so is how you systematically turn the information into classification data. You have potently delivered a result showing that non-clicking data, though often underrepresented by both platform algorithms and academic data scraping, have nearly the same weight of importance as clicking data. That being said, I have two concerns/questions: 1) You pointed as an example that the currently prevailing algorithms that often promote clicked contents and neglect non-click ones would exert some significant biases, failing to distribute posts that are meaningful. But as you noted, such non-clicking behavior occur since the audience find it inappropriate to respond to the more lengthy contents with a simple click. I feel that such algorithms are in the interest of both the modern internet audience, who increasingly prefer small, fragmented pieces of contents to a larger chunk, and the social media platforms which cater to the audience's appetite. 2) You highlighted how vital information can be lost if we only use clicked data in researches. I wonder how social media platforms can improve their api to make non-clicking information available? Also would it be in their favor to do so since the platforms might wish to keep it themselves for business purposes?

[afchao]

Thank you for speaking with us! In your paper, you reference other social media platforms and the general nature of social media; however, by the very nature of your study, your findings are limited to facebook for the time being. Do you expect to see similar patterns of use across most social media platforms? Of specific interest to testing the generalizability of your analysis are A): ones for which the user is kept relatively anonymous (and therefore may have fewer factors constraining their click behavior) and B): ones for which the user may incur greater-than-usual consequences for inappropriate behavior (e.g. Linkedin, if anyone still uses it...) - in general, examining click/non-click behavior in other contexts seems like the natural progression of this program, although other members of this program have proposed some interesting extensions here as well!

Thanks again!

[Yutong0828]

Thanks for this very refreshing work! I have never thought that clearly about why people click or not when they are using social media. I have two questions for you. First, it's impressive that you combined qualitative interview with eye-track method, which added more information to the results, but I feel hard to reach solid conclusions with paragraphs extracted from different interviews, as we can't tell if a person's answer is representative or not. So I was wondering if we could analyze the interviews quantitatively, and look for similar patterns, in order to come up with more convincing explanations for the findings. Another question is, one of the findings in the paper is that non-click behaviors don't necessarily signal inactive or less attention, but may reflects intentional or deliberate thoughts. However, some participants mentioned that though they won't click "Likes" under the message, they may still contact the person privately if they feel necessary. In this way, they don't click but are also engaging actively. I am kind of confused that why contacting person privately is not a form of "click" behavior, as it also requires people to do something. I think if we broaden the definition of "click" behavior, it might be easier for us to distinguish active and non-active social media users through this indicators. Thanks very much for sharing your great work!

[anqi-hu]

Thank you for sharing your work with us and your newly-proposed typologies seem really interesting in challenging the traditionally used dichotomy of user behavior. I have two questions:

1. While reading through the procedure of your experiment, you had the subjects browse their FB pages on desktops. Eye-tracking technological concerns aside, I was wondering if changing the device on which individuals scroll and "click" would mean different for their browsing behavior (as initiating interaction would become handier) and shed a different light on the findings. Do you think switching to tablets or mobile devices would perhaps alter the observations?
2. You mentioned in your results that the amount of social content one has in their feed was predictive of their clicking behavior. While that makes sense, I think that the social distance between the viewer and the posters could be important in terms of predictability. For example, individuals might habitually like or comment on their best friends' updates, regardless of the content; they could also withold such interactions with those they are more socially distant from because of it. Do you think social distance should be incorporated in future studies of the clicking behavior? How would you go about measuring this distance?

[bowen-w-zheng]

Thank you for this interesting work!

I was also thinking about potential ways to scale up the study and to run the study remotely. There are some attempts to mimic eye-tracking:

• BubbleView: Participants are presented with blurred images (mimicking peripheral vision) and they can click on the images to view a small patch of the image (mimicking central vision). The clicking patterns are taken as a proxy of eye-movement/attention allocation sequence.
• Mouse movements also correlate with eye-movements and some studies suggest mouse trajectory can capture unconscious mental mechanism.

Do you think these measures could approximate eye-movements? If not, what crucial information is lost when using these measures as proxies?

[wanitchayap]

Thank you for your work and presentation! I have a small question about your sample choice. You mentioned that the study excluded emerging adults because this group over-represented in the literature. However, shouldn't we still include this group in the sample if we want to understand the non-clicks in general? Why did you opt to exclude this group entirely instead of making sure the portion of emerging adults in the sample reflects the actual proportion of emerging adults in the population? (Or alternatively, we can even consider using MrP if needed?)

[siruizhou]

Thank you for sharing this thoughtful analysis. I've never thought of viewing and clicking as two independent behaviors when using social media so I think it's more meaningful to study the overall effect. I wonder if you also find clicking patterns related to the number of interactions already associated with the posts before people click? I think people tend to 'follow the crowd' and might change their typical behaviors.

[hhx2207061197]

Thanks for sharing. I think the social distance between the audience and the poster may be important in terms of predictability. Do you think social distancing should be included in future research on click behavior?

[linghui-wu]

Thanks for sharing this exciting piece of work! I have a similar question with @siruizhou in terms of how the echo chamber enhances individuals' certain viewing and clicking behavior. In other words, how the research design considers the underlying impacts of Facebook algorithm confounders?

[yiq029]

Thank you for sharing this interesting paper! The connection between passive and active interactions on social media is very inspiring. I am wondering if we choose research method like interview, what the sample size is supposed to be to avoid the misrepresentation as much as possible? Or, how the sample size is determined? Thank you~

[yierrr]

Thanks for such an intriguing paper! I have two questions about the paper: first, since the subjects were asked to look at their own social media, how did you make sure that they didn't do the same before the experiments, i.e., that what they were looking at was new to them? Because I think if someone has seen some posts before and encounter them again, they may pay some attention as they might recall seeing them, but not click as they know they have seen the content. Second, I find it kind of counterintuitive that liking to please others predicted fewer clicks. If someone is really trying to please someone else, wouldn't they actually try to be more interactive to show how much attention they are paying and to maybe get noticed? Thank you!

[YuxinNg]

Thanks for sharing this interesting paper. My question is during different time, a person may act differently. For example, the same person may act (viewing and clicking behavior) differently in the morning on his/her way to work and when he/she is lying comfortably in the coach after work. Also, I agree with @hhx2207061197 that pandemic/ quarantine/ social distancing may also change people viewing and clicking behavior. Do we need to take all these factors into consideration? Thanks!

[NikkiTing]

Thanks for sharing your work! I really look forward to your presentation during the workshop. As Facebook is adding more features, I think more people are also using the platform for entertainment purposes or as a marketplace. Given this, could an individuals' primary purpose for using Facebook (e.g., someone might only be looking for good memes, videos, or things to buy and not care about the audience) also be an important factor determining their viewing and clicking behavior?

[Raychanan]

Hi Dr. Ellison, Tiktok uses an algorithm to suggest videos to users that they might like. However, I have a problem: if a user gets bored of watching videos about pets (when in fact he/she is still interested in pet-related content), he/she may click on the "reduce recommendations" option. How do you solve this paradox: the user is always interested in pet videos, but the algorithm is not advanced enough to make him or her click "dislike".

[vinsonyz]

Hi Nicole! Thanks for sharing your paper with us. Since the sample is small, I wonder if we can take advantage of the observational data to study the non-clicked context.

[minminfly68]

Thanks for sharing with us this paper. I have a similar question on sample size. It was mentioned that the emerging adults were excluded from the study, can you elaborate more on that? Many thanks.

[YileC928]

Thanks for sharing! It is a really interesting angle to approach the topic by exploring the "non-click". My question is, how did you construct and balance the three methodologies (eye-tracking, survey, and interview) to let them complement each other? Besides, as many classmates have already mentioned, I'm also quite curious about ways to scale up the research. Thanks in advance!

[william-wei-zhu]

Thanks for the paper. What will be the long term implications of this research project? Should social media companies prioritize "reading time" over "clicks" as a better predictor for interest in their recommendation algorithms?

[wu-yt]

Thanks for the interesting research! I’m wondering how would you consider the impact of your results on morals? Can your research be exploited in a negative way? Thank you!

[lihanhuisherry]

Thanks so much for sharing your research with us! I am curious that, due to the difficulty of tracking "non-click" behaviors, how could we differentiate the intentional "non-click" behaviors from "did-not-notice"? Thanks!

[Yilun0221]

Thank you for the presentation! It is really a fantastic research about social media and my first time to see reseasrch about 'click'. My question is, does the data about people’s click behavior violate the privacy of users?

[k-partha]

Thanks for the paper. It would be fascinating to reverse engineer the social signalling decision mechanisms responsible for clicks/non-clicks. Do you think these results shine any light on a generalized theory of the signal-incentive mechanisms underlying clicks? Do you think it is possible to do this given ethical considerations on processing private feed data?

[chiayunc]

Thank you for the wonderful research. I have been observing my own non-click behaviors ever since I get the overwhelming feeling that what I do and not do on social media will eventually make its way back to haunt me. So for me, the cause that made me start being extra reserved on social media was the knowledge that what I do would affect what I get. I am really curious then, for the third category of the non-clickers, do you see any changes before and after the incident of for example, Cambridge Analytica? or other major news outbursts that raise the awareness of such issues? Thank you.

[XinSu6]

Thank you so much for the paper. I have been wondering for how my moves on social media like clicks or non-click can be stored and analyzed. My question is how can further application be conducted? After study the users' behavior, what kind of possible applications can there be? Can this be potential used to manipulate users' behavior or try to change their mind on things? Thank you.

[MengChenC]

Thank you for your great work. The research incorporates objective eye-tracking and subjective survey data to validate the method. In order to mitigate the observer effect and expand the sample size, have you considered the research design based on always-on Facebook big data? I am curious about the possibility and difficulty of using already-made data in this research. Thank you.

[TwoCentimetre]

Thank you, I should say that this is a really interesting and thought provoking topic. For one thing, before I read this paper, I used to think that a computational social science research requires at least hundreds of thousands of pieces of data to be called a computational research. But this paper only has 42 participants and use a really traditional research design to do the experiment. So, I am confused why this paper is considered as a computational research. For another, on most social media, there is only a button for you to express "likes" but no place to say "dislike". I mean when someone post pictures and I don't like them, there normally is not a thumb down button for me to click. Is this could be something that matters in this research?

[heathercchen]

Thank you for presenting us with this interesting topic! Given the arguments in your paper, computer scientists should put much more emphasis on "non-clicking" behavior since they are not so "passive" as we used to assume. However, according to my understandings, the reason why we are more focused on clicking is that it is easier to track than non-clicking. Collection of the times that one user clicks on a certain program or website can be done by a simple function, while non-clicking data collection requires techniques that can only be conducted within a laboratory. What do you think of the practical limitations? Thanks!

[WMhYang]

Thank you very much for sharing this paper. The idea to use eye tracking to identify the viewing and clicking behavior is fascinating to me. However, as others have pointed it out, it is difficult to scale the data size up given the limitations of it. Hence, I was wondering if we could utilize some algorithms or machine learning techniques to mimic the viewing behavior of individuals to enhance the data availability and size based on your findings. Thanks.

[ydeng117]

Thank you for such an interesting presentation! As many of my colleagues have pointed out the data size of this research is a limitation. It totally makes sense to me that the research applies a qualitative method, as we have to know about how participants make sense of their browsing behavior to separate the clicking and non-clicking behavior. I am curious that, base on the finds in this research, could we develop a quantitative research design to study a larger population?

[ginxzheng]

Thanks for sharing! It's very impressive to apply eye-tracking methods on Facebook behaviors. One question is, why do you only consider "click" versus "non-click"? (given you include "like", "comment", "share", etc. all into "click"). To my knowledge, there could be many more nuances between such behaviors, and the questions you raised may be sliced into many different sub-questions. Also, I wonder for the interview analysis, would it be possible to enlarge the participants' size, using any outsourcing digital methods? Thank you!

[JuneZzj]

Thank you for presenting in advance. It is interesting to know in social media, how people behave in such a way and which kind of behavior can be treated as "active". I noticed that in the paper, you mentioned something about the fact that non-click can be a useful concept that helps researchers learn visibility management. By hiding their activities, how researchers investigate users' behavior of determining their visibility in social networks.

[chentian418]

It's really interesting to know how “non-click” plays an important role in contemporary online communication environments! I have a question about the unbiasedness of the data. While the data is collected in a digital lab setting, I was concerned that the university staff member would be a featured population regarding their viewing and clicking habits of Facebook. Moreover, although oversampling could adjust the under-representation of men and people-in-color, how did you decide this oversampling in recruitment, in a way that wouldn't impact the unbiasedness of both the sample population and the related viewing & clicking indicators? Also, would there be any exaggeration of the effect captured in such lab experiment setting? Thanks!

[mingtao-gao]

Thank you for sharing this paper. It's a very interesting topic to me since "non-clicking" seems to not be in the mainstream of research. However, I do get confused on the operationalization of clicking. Clicking is considered as "an action on Facebook that interacts with a post and generates a visible trace to other users", which seems to me is the same as social interactions on social media platforms. There are many intentions behind a click besides social interactions. Users may click on photos but not click on any buttons to engage with other users. Users may also click into one's personal page without any interactions left. Personally speaking, the definition of "clicking" here may be too narrow by focusing only on the interactions.

[Dxu1]

Thank you for sharing this interesting paper. I have the following thoughts:

1. I am interested in if the mood of the participants, something that seems outside the data collected by the study, could have an effect on their (non)clicking behaviour. a. How different would the participants behave if they were not under seven-minute session but in a more casual setting, that may make them feel more "comfortable"? b. Speaking totaly from personal perspective, but I seem to click more when I am in a better mood and do not click at all when I am in a low mood. From your interview, do you get a sense that mood might alter their clicking behaviour?
2. Related to previous posts, I am also very interested in scaling up the study, to study especially non-click behaviours. How could we get "non-click" behaviour from digital records?

[FrederickZhengHe]

This is a very attractive paper since past literature seldom focused on non-clicks because clicks surely expose more explicit information of the users. Also, it discusses four groups of people: indiscriminate clicker, engaged clicker, engaged lurker, unengaged lurker. So my first question is: are there any quantified criteria or thresholds to measure whether a given person belongs to one of the four groups? My second question is: is there any correlation between the number of visible previous clicks and a person's click or non-click behavior? For example, if a Facebook post has got more than 3,000 clicks (thumbs-up, sad-face, angry-face, etc. combined) by the moment a person has noticed it, would he/she be more likely to click on this post? Also, what if this post has got only 20-some or 30-some clicks? Thanks!

[hesongrun]

Thanks a lot for your presentation! This is very interesting research. How can we apply this research methodology to other disciplines in sociology? Thanks!

[kthomas14]

Hi Dr. Ellison, thank you for sharing your research. I must say, the paper was very easy to read and the breakdown of each user type was very interesting. I was wondering if you had any plans to expand this research design to platforms like twitter? While I can only speak for my own feed, it seems like people don't share as much personal information on Twitter and are more prone to interact with people they do not know than they would on a platform like Facebook. It would also be interesting to track user interaction with others who are exclusively "internet friends" compared to "real-life friends", where "online friends" are not geographically close, but users spend a lot of time interacting online.

[jsoll1]

Thanks for sharing your paper with us! It's definitely not directly useful for me in the capacity of a Facebook user, but it's still really fascinating to think about. Thinking about the non-click being significant is super cool because there have been times in which I purposely didn't click for social reasons. My question is what effects do you think people shifting between the four different conceptual categories of social media use can have on them?