Social Computing 2

From CS2610 Fall 2014
Jump to: navigation, search




Reading Critiques

yubo feng 5:13:18 11/11/2014

“Design Lessons from the Fastest Q&A Site in the West” by Mamykina et. al. is a paper that discussed different styles of question and answer sites, in particular Stack Overflow(SO). They explored the qualities of SO that has made it so successful in just a couple years. They did by using a mixed method approach, first analyzing usage data from the site, then secondly interviewing users, designers, and founders of the site. From the usage dataset the authors categorized four types of users; active, shooting stars, low activity, and no activity. They then explained the distribution of posts among these user groups, and the distribution of the users in to the groups also. They found most of the users are low activity or no activity, but most of the questions and answers come from the active users, almost 30%. Then through interviews the researchers discovered three things they think made SO successful. They made competition productive by creating a finely toned reputation and badge system to create extrinsic rewards from users to answer questions, this was very effective because the average time until as answer is posted is 11 minutes. The next factor is credibility in the community. The founders were able to easily achieve critical mass because both were extremely active in the community and wrote successful blogs. Both has followings that could easily be transferred over to their new project, SO. Also, because the community already existed prior to SO, there was great trust between users, thus resulting in better quality questions and answers. The final factor is their evolutionary approach to design. Instead of sticking to a short bursts of user feedback, as suggested by the iterative design cycle, the designers are in constant communication with the community and pushing updates daily. They designers also used the production site to experiment with new designs, also like SO is a perpetual beta to be improved in accordance with the communities wishes. I found the user interviews to be more insightful than the data analysis. The all data analysis mostly showed that SO follows the characteristics of other social computing systems. I thought the user interviews exposed more about how SO became so successful and proved to be more inspirational. “Predicting Tie Strength With Social Media” by Gilbert and Karahalios is a paper that details the design and predictability of a “How strong?” model used to predict the tie strength among Facebook friends. Tie strength is closeness of two people. One may have a strong tie with family while having a weak tie with a new co-worker. There are well established categories for tie strength predictors, which were used in this paper’s model, such as intensity, intimacy, duration, reciprocal service, structural, emotional support and social distance. The experiment performed was to ask subjects to answer a few questions about a friend, to provide true data for the predictor to be tested against, while collecting other features about the subject’s relationship to that Facebook friend. Some of these predictor features include wall words exchanged, wall intimacy words, wall & inbox emotion words, links exchanged on wall, and education. The model was able to predict the tie strength with great accuracy, within one point of the subject’s true rating. The model also showed the contribution of each category of tie features. intimacy, intensity and duration contributed the most 31%, 20%, 17%, respectively, while other categories like emotional support and services were much less important. This paper showed a very interesting application of machine learning to CSCW system to learn more about group dynamics, specifically tie strength.

Wenchen Wang 16:28:08 11/11/2014

<Predicting Tie Strength With Social Media> <Summary> This paper introduces a model to predict people’s social tie strength based on social media, Facebook. <Paper Review>Tie strength is a combination of time, emotional intensity and intimacy among people’s social community. There are seven dimensions to measure tie strength. The researchers asked two questions before experiments. One is can social media predict the tie strength. The second question is what are the limitations of tie strength prediction based only on social media. Based on the seven dimensions, the researchers choose 74 Facebook variables to predict tie strength. The result shows that intimacy makes the most contributions to the prediction. And they also found out social media can predict 80% relationships. This number could answer the two questions researchers asked above. Social media can predict tie strength, but it cannot predict all, which means there are limitations. So just relying on social media cannot predict all relationships. Researchers found out other relationships by interviewing the participants. And they complemented the social media and gave some suggestions to help social media to be more precise. <Design Lessons from the Fastest Q&A Site in the West><Summary> This paper discusses about the factors behind the success of Stack Overflow, which is a very popular Q&A website. <Paper Review> There are three major factors of the success of SO. First is making competition productive. SO relies heavily on community moderation. Active participants can vote questions and answers of others up and down. They are also able to close inappropriate question. Second factor is the credibility in the community. Even before SO, the founders had a combined readership of about 140000 people. SO has initial critical mass of users, which is a key point of social software system. The third factor is evolutionary approach to design. The SO design team adjust the design of the site and update the modification very often, which make their design iteration fast. In addition, designers have tight feedback loop with users. To HCI prospective, analysis the pattern of user interaction of SO could also help HCI researchers know more about common user groups and behaviors.

Wei Guo 22:28:17 11/12/2014

Reading Critique The first paper, Predicting Tie Strength with Social Media is about using a predictive model to bridge the gap between theory of tie strength and data from social media, and prove that the data from social media successfully predict the tie strength. The paper first introduces the tie strength as the measurement of relationships. By using seven dimensions of tie strength, the authors predicted the friendships in the Facebook dataset. The authors then developed five questions to let users rate their tie strength. The results show that social media can predict tie strength. There is a very good evaluation example in this paper. From a dataset, the authors used the calculated predictions results of the dataset as independent variables, and let users’ assess of tie strength as dependent variable. They perform the relationship in a very convincing way. There are a lot of control variables to largely avoid the faulty of results. The analysis of the results is very understandable and convincing. The author not only explain why some error predictions prompted out, but also give the examples. The second paper, Design Lessons from the Fastest Q&A Site in the West, is about analyzing Stack Overflow, a very successful Q&A systems, to understand its success. The success due to an a priori superior technical design, high visibility and daily involvement of design team within the community they serve. Clearly, the main goal of this paper is to understand the success of Stack Overflow. The authors in this paper conducted a statistical data analysis, investigated answer time, user types, suitability for different question types…. They analyzed the results and compared it with several other not that successful Q&A systems, and then draw a conclusion that: Making competition productive, credibility in the community, and evolutionary approach to design help to make Stack Overflow successful; if any system wants to copy the succeed, defining clear topic boundaries, demonstrating commitment to the project, and showing viability through sufficient user activity in the test phase might help. Although this paper does not as interesting as the first one to me, the ways it uses to analysis and interpret data are still outstanding.

Nick Katsipoulakis 23:29:19 11/12/2014

Predicting tie strength with social media :: This article presents a quantitative study about the significance of tie's strength in social networks. The authors generalize the argument (through an extensive user-study) that a tie's strength is a very important factor that affects user behavior in social networks. Previous work on the same field has failed to measure the strength of ties in depth. However, the boom of social networks that we are experiencing the last decade, has brought this matter to the attention of researchers. By examining user behavior the authors succeed in gathering data about the effect of social ties in social networks and manage to break it down in different factors. Also, the authors manage to build a prediction model for analyzing user behavior based on a tie's strength. This paper is easy to read and the authors' motivation is clear from the beginning. "Social Mining" is a very important field that affects many research areas of Computer Science. //------------------------END OF FIRST CRITIQUE ------------------------------------/// Design Lessons from the Fastest Q&A Site in the West :: This article discusses social phenomena and the evolution of one of the most popular Q&A sites for developers, StackOverflow. The authors present statistical data gathered over 2 years of study, and results/conclusions made through interacting with community members of StackOverflow. Personally, I am a member of StackOverflow for three years now, and most of the social behaviors explained in this article are familiar to me. As StackOverflow is one of the pioneeers in this type of community sites, the authors felt compelled to study the user behavior patterns and the incentives behind its design. From the point of view of the creators of StackOverflow, the authors reveal their intentions and how those affected the websites philosophy. Also, opinions and behaviors of active users are presented through statistical analysis of data and through interviews with the users themselves. Even though I did not understand the reason for presenting the statistical data in so much detail, the contribution of this paper is clear to me. Several behavior patterns have been demonstrated and the methods the creators of StackOverflow used to keep the user community active.

Qihang Chen 0:19:23 11/13/2014

Predicting Tie Strength With Social Media This paper presents a predicting modal which is used to solve the gap between practice and theory on the issue of tie strength with social media. This paper has the point that social media are structured under the relationships. And social media acts exactly on different impacts brought by different relationships. Different ties strength plays different functions. For example, weak ties are known as being good for job hunting while strong ties are good for health care of a family. However, theses assumptions were built based on theories. This paper then investigated the real cases using social media data. The paper chose Facebook as the testbed which is the most popular social network nowadays. Seven dimensions were introduced here (Intensity, Intimacy, Duration, Reciprocal Services, Structural, Emotional Support and Social Distance) under four proxies: possessing at least one mutual friend, recency of communication, interaction frequency, communication reciprocity. The paper then carried out studies to answer two questions: could the seven dimensions predict the tie strength and what the libations are behind the tie strength modal. The studies were very straightforward. All the data could be summarized from the Facebook of the participants while the participants could rate their friendships as the comparison. Some skills about composing investigation questions were introduced. For example, To learn how strength the tie is, we do not need to ask the participants directly. We could split the question into concrete situation such as whether they feel comfortable to ask that friend to loan money them. The result shows that the social media is able to predict tie strength. But there are some limitations about the study settings. For example, the emotional support is hard to measure using quantify methods. The paper is a nice example to show to how conduct an social computing related experiments to learn some known assumptions, especially the way the authors design the experiments and analyze the data. Design Lessons from the Fasted Q&A Site in the West This paper analyzed a QA site carefully and design new approached to improve on the utility and performance of that system for technical domain. Basically, the aim of the study including interview is to find out the factors that leads to a successful Q and A site (Stack Overflow). The first question is that how to understand the factors behind the success. The paper answered the question by conducting a statistical data analysis of the SO corpus to find out the usage patterns. Specifialy, the paper looked into answer time, suitability for different question types, and possible extensions of the SO model to other domains. The result reveal that there are several key factors: fast answer times and high answer quality and a strict set of community guidelines, informational answers. Meanwhile, these factors are resulted by the design strategy of the founder of the website. Basically, some key ideas like making competition productive, building credibility system in the community, evolutionary design approached make the website powerful. The paper then introduced the data collection and analyze process in detail. It is important to learn how the studies could be composed systematically and how to find the key factors that have big impacts. The very first thing is to learn how the website works. Also, it is important to statistically analyze the results or the achievement of that website from the website's data. And what the data means. Based on the findings of the patterns, the paper then introduced the quality study with the cofounders, design teams to find out what are the factors that lead the patterns. It is informative while reading this paper. The paper shows the importance of the ability of analyze as researchers. We need to learn how to form the question, how to systematically review, categorize and analyze the data. The key idea of this paper is to find out the factors behind the success of the website, as well as the challenges, which are very useful for future designs

Qiao Zhang 0:20:09 11/13/2014

Predicting Tie Strength With Social Media In this paper, a predictive model that maps social media data to tie strength is presented. The model distinguishes between strong and weak ties. The author uses several selected parameters to predict how strong the social tie is, predictive variables including intimacy, intensity, duration, emotional support etc. The result turns out to be very accurate with over 85% accuracy. For the friendships they had the most difficulty predicting, the authors conducted follow-up interviews to understand the reasons. Strong ties are the people you really trust, share the same social circles and most like you; weak ties, conversely, are merely acquaintances that provide access to novel information that not circulating in the closely knit network of strong ties. One uniqueness of this work is the leverage of social media. Participants don't have to recall, such that the problem of retrospective informant accuracy is avoided. The rich information of friend lists and interaction histories can also be utilized. It can also potentially benefit the users of social media, too. A lot of independent variables are selected as independent variables for the black box. Variables are classified into 7 categories: intensity, intimacy, duration, reciprocal, structural, emotional support and social distance. The dependent variables are 5 tie strength questions, with each one variable of continuum levels. A linear model is used to predict the relationship between the independent variables and one of the dependent variables. An interesting point made in this paper is the explanation of why the model does not fit the last three questions as well as the first two; the authors infer that it might be resulted from participant fatigue. Randomizing can be a solution, but it may also annoy the participants. Hence the most interested question is prioritized. Another good point is that in order to understand the limitations, the authors conduct follow-up interviews about the friendships they had the most difficulty predicting. They pick the friends with the highest residuals and ask the participants about their relationship. The reasons are quite interesting to me; somehow they reflect the intrinsic complexity of human emotions. But as can be seen from the examples, the model tend to underestimate how strong the tie is. I think it implies that good friends do not necessarily interact in social media as intimate as in real life, which applies to me, too. This work seems to be interdisciplinary with social sciences. I am wondering how computer scientists can collaborate with social scientists to get the most of the abundant data generated everyday on social networks. ==================================== Design Lessons from the Fastest Q&A Site in the West This paper describes a popular Q&A site, Stack Overflow and analyzes factors in the site’s design and evolution that contributed to its success. The paper first gives a statistical data analysis of the entire SO corpus to understand usage patterns. The authors investigated answer time, user types, suitability for different question types, and possible extensions of the SO model to other domains. To ground this aggregate view in concrete user experiences, the authors also conducted a qualitative interview study with users and the design team. They believe that three factors are the main reasons SO succeeded: 1) Making competition productive; 2) Credibility in the community; 3) Evolutionary approach to design. Analyzing social networks sounds very interesting; however, I always have the concern of this kind of research: the process seems to be having the data first, and trying to get some insights out of it; not really like the "traditional" way to research that finding the question first and then collecting and analyzing the data. When got the raw data, most researchers will do simple statistics on it just like the authors did in this paper, to get the overview of the dataset; and if they see anything interesting, they try to dig deeper into it. Somehow I feel like this approach is not very straightforward to me. As seen in this paper, a dozen of statistics are presented to answer the question of "how well does stack overflow perform". From the total number of users, to answering time, to different types of users, the paper illustrates the overview of the SO dataset with the help of visual plots. Many of the plots exhibit power law distribution (big-tail) as expected as an online social networks. To better understand the driving factors behind these patterns the authors conducted a qualitative study of the community. They did interviews on the users, site designers and founders of SO. Three reasons contribute to SO's success: 1) founders’ tight involvement with the community, 2) highly responsive and iterative approach to design, and 3) a system of incentives that promoted desirable user behavior.

phuongpham 0:37:22 11/13/2014

Predicting tie strength with social media: this paper presents a study on tie strength using data from Facebook. This work is interesting to me because Facebook seems to be a myth compared to Twitter of which data can be partially downloaded from an official API. The authors have found a way to collect data from Facebook and used the power of this large social network for this study. Following the style of papers we have read, I think this paper is a good one. First, the paper has a clear statement about its research question which addresses both theoritical and pratical concerns. Second, the authors have done a depth analysis section which sheds a light for future research. However, there are some concerns about the work. I am not sure how can a script can collection all interactions between 2 user on Facebook. As the authors claimed that they collect data at client side while all interaction history may be only available at Facebook's servers. Another point is the scale of this study. Given each user rate 63.4 friends, will the ratings accurate with such a large number? What if we want to get ratings from all friends? Again, the authors have done a really good job when getting research information from Facebook. I just wonder how we can scale up this project. About the feature, there would be another important feature which can take into account is the ratio that a user spend for a friend. I happen to know someone who register Facebook just to follow and discuss with a small number of people, e.g. parents connect with children only. The intuition would be if you spend most of the time for a friend, that is most likely a strong tie. Last but not least, the tie strength is bidirectional. A may strongly connects with B but B may not, e.g. celebrities and big fans. This would be another interesting research question. ***Design Lessons from the Fastest Q&A site in the West: this is another paper about CSCW systems. In this paper, the authors also mentioned about some indicators for a successful CSCW, such as critical mass user. However, what I found interesting in the paper is that the authors have interviewed the system's founders and designers. This gives another feedback direction to analysis the system. Compared to the previous paper, we only see 1 way of feedback: users' opinions. This suggests a new way to do research on CSCW where we can get analysis information from both parties: users and producers. The paper also gave helpful information about the successes and challenges of SO as well as some examples of fails when cloning the model. I still have a little confusions about the difference between human-centered design process and this SO building process.

Mengsi Lou 2:14:28 11/13/2014

Design Lessons from the Fastest Q&A Site in the West ---------This paper discusses the successful case of the Q&A site Stack Overflow and mainly analyzes factors in the site’s design and evolution that contributed to its success. ---------The Stack Overflow is familiar for us programmers. The main feature of Q&A site is fast answer times and high answer quality. And in the meantime, this feature also means the site is strongly and publicly involved in both control of and debate within the community. And the author proposals three main factors that makes the Stack Overflow. First, making competition productive. There is a voting system that works for the answers and the voting is available for all the users. Thus the competition led to the intense short participation for some users, and long sustained participation for others. The second factor is credibility in the community. The thought-leader status and visibility contributes to having a lot of dedicated users. The third factor is evolutionary approach to design. They have the continuous feedback loop with their users so that they can know the challenges and concerns of their users and modify their feature better. ---------The analysis part the author uses the method Structural Analyses Capture Aggregate Use, and also make compare with the Yahoo answers. The author states that the Stack Overflow is smaller size than the general sites, but the participation is more frequent than general. The amazing data is 92.6% of questions are answered and most of them are answered more than once. And also the answer is very fast after the question posted. The mean value of the first answer is 11 minutes and the accepted answer is 21 minutes. The participants are important that can be seen from frequent users post more answers than questions. That makes the activities alive. ---------The above part mainly discusses question patterns and the performance of Stack Overflow. Next the author talks about the design points that makes Stack Overflow successful. First, they adopt a rapid prototyping approach driven by direct and immediate user feedback, which makes the users tight with the feedback. Second, the Rapid Design Iterations. Through rapid iterations they can find the problem quickly and also update the site frequently. ---------This paper inspires me for the social media design that should design the site features target on the users’ need and behavior. And also a successful site need new and flexible ways of design that make the social networking more efficient.

zhong zhuang 2:19:13 11/13/2014

In this paper, the author discussed about the most popular Q&A site for programmer – I believe every programmer, no matter what programming language he is using, has visited this website. The paper reveals the secret behind the success of SO. First it analyses the patterns of user interactions with the site. Most users are passive, they only look for answers, only a few users are very active, less than 5%. But most questions are answered by these active users. The author compares SO with other Q&A websites such as Yahoo!Answer, we should note that Yahoo!Answer is a subjective QA website, that means many questions doesn’t have an answer or users are not seeking for one specific answer, they are mostly asking for advice or comments. So anyone can answer questions at Yahoo!Answer, but SO is an objective QA website. All questions do have a specific answer. Not everyone have that specific knowledge to answer the question. But SO is also one of the largest QA website. Then the author conducted various user studies to show the reasons for the site’s success, author interviewed the founder of the website, core design team member and regular users. Then the author elaborated some highlights. First is the founder’s tight involvement with the community, second is highly responsive and iterative approach to design. The last one is a system of incentives that promoted desirable user behavior.

Xiaoyu Ge 2:51:31 11/13/2014

Predicting Tie Strength With Social Media This paper presented a predictive model to map social media data to tie strength. The result shows over 85% accuracy to distinguishing between strong and weak ties. The paper concluded that modeling tie strength could be use to improve social media design elements, including privacy controls, message routing, friend introductions and information prioritization. With the prediction of ties, the social media software can develop more specific functionalities to fulfill user`s need. And since the accuracy of the model introduced about 85%, it is high and can be actually applied in to industry. It is true that social media treat everybody the same, and user himself manually configured all specific features to treat their friends differently nowadays. However, if implemented an auto configuration setting by using the tie model, the user themselves might not really want to use the default one, they will need to reset the settings since the relationship with people changes and the accuracy is not 100% right since there are still many aspect of influences of ties that failed to considered in this model. It will be better if the model can be auto adjust such as adding additional influential conditions. The concept introduced by the author is useful, but still it need improvement to achieve better result. Design Lessons from the Fastest Q&A Site in the West This paper introduced a popular Q&A sit for programmer and software engineers, Stack Overflow and analyzes factors in the sit`s design and evolution. Analysis of the patterns of user interactions with the site helped to highlight some of the more common user groups and behaviors. Qualitative study focused on findings, credibility in the community and challenges. There are three reasons make the website successful. Firstly, the website founders’ tight involvement with the community. Secondly, it is highly responsive and iterative approach to design, and thirdly it is a system of incentives that promoted desirable user behavior. It is true that stackoverflow is a successful website. There are not many website discussing technique problems people encounter in such detail and the questions were answers and covered a lot of ranges. It is an innovated website since the founder really understand the product. Because the website should be able to help answer questions to build up it`s user groups. And since it is for technique people it is need to be right to the point, and the most useful answer should be through out to user`s face and present in a simple way. I agree with the paper`s method of evaluating and analyze the paper.

SenhuaChang 3:00:11 11/13/2014

Predicting Tie Strength with Social Media The authors give a linear model about how to calculate the tie strength in social media. While I know lots of researchers from complex network, data mining, machine learning areas also do some learning and mining in this social data (they focus on how to construct the social graph and then apply learning methods on the graph model), the authors in this paper extend the dimension of the data (Based on Facebook, the authors put a deep emphasis on the social media and identify 74 variables as potential predictors). While I admit that this is a novel way to calculate the tie strength from social media, but since this paper is published on CHI, I think the authors should use as least the same length of paragraphs to illustrate how these tie strength and other observations are helpful to improve social media design element instead of talking this briefly in the discussion part. Another question is that I think if the authors can separate the degree of tie strength into multi-categories, this will more meaningful. Design Lessons from the Fastest Q&A site in the West The authors make a deep analysis about a popular Q&A site Stack Overflow. The authors speak highly of the technical design, high visibility, and involvement of this Q&A site. I think the entire analyzing part can be divided into two parts: user pattern analysis and the success reasons. I think one thing we can learn is the analysis technique about the user pattern analysis part. Those data in Stack Overflow can be viewed as a time series dataset. The authors choose different perspectives to arrange those data and get lots of meaningful observation and finally use intuitive figures to illustrate them. We can apply this to many other time series data set. For the second part, the authors owe the success to those three reasons: making competition productive; building on exiting credibility within the community; and adopting a continuous evolutionary approach to design. And the authors use lots of paragraphs to explain these reasons. However, in my opinion, I do not think the second successful reason can be applied to other application. Since the website is about programming Q&A and the author has some programming background, so they make some creditable interactions. It’s difficult to generalize. But we can switch to invite some famous person in that area instead, which will compensate the creditable interaction.

Bhavin Modi 3:05:27 11/13/2014

Reading Critique on Predicting Tie Strength with Social Media The paper is about trying to predict social models, ties and relationships in this case using a prediction algorithm. This algorithm takes into account many variables using existing research and converts them in terms of social media relevance. We move on from the traditional HCI papers, we are focusing on trying to predict behavioural patterns from data from social media like Facebook. This is an interesting and a very important field of discussion. Many of the major companies invest millions for the same purpose and this is one of the major reason why Facebook bought WhatsApp for $19 Billion. Collaborative prediction used by amazon talked about in previous papers and lectures follows the same pattern. To have an idea about how users think and the social patterns, is vital for the success of any application or organization. The main question asked was if the existing literature suggesting the seven dimensions to determine a tie Intensity, Intimacy, Duration, Reciprocal Services, Structural, Emotional Support and Social Distance are enough in context of prediction with the use of social media and the limitations of such an approach. This is not an easy process, as observed from the author’s use of 74 Facebook variables. The prediction is 87% correct, which is impressive, given the complex nature of relationships and variability involved. Having such a model indeed has many practical implications especially in terms of privacy and control, but it loses out on unexpected behaviour of people and some new connections forming. Another flaw according to me is that initially the users were asked to rate some their friends randomly and this data was used to predict the accuracy of their model. This is one sided view of the tie between two people, one may consider his tie strong and the other may not. This is a discrepancy unless data is collected from both sides. This being said, careful effort has been done to list the problem that they had only done this for Facebook and the future work in details. Overall I find the paper decisively well written with the clear diagrams especially the predictive power of the seven tie strength, How strong? Model. Bottom Line we have read about hoe to predict ties from the data basically left behind by people so as best help them by gathering relevant information about the social structure of their life. -------------------------------------------------------------------------------------------------------- Reading Critique on Design Lessons from the Fastest Q&A Site in the West The paper analyses Stack Overflow, a Q&A site and finds out the reasons for its success compared to other such sites. The author’s motivation is to find the design principles that make Stack Overflow one of the best sores for technical queries and what effects that has for HCI and CSCW researches. To highlight the effects is that SO uses an iterative prototyping design with continuous feedback from the users, which is one of the reasons for its success. But to implement this in an application oriented environment is one of the major challenges for researchers. Following from knowledge from previous papers, which discussed crowdsourcing that providing rewards and making it like a game, providing a competitive feel, makes use engaging. SO use of the voting and the reputation system provides this incentive. This paper in a way feels like it combines all the previous papers read to create a great platform and as a result outperforms the rest. Remember the groupware paper discussion on initial critical mass, here an important factor is that the founders already had a substantial following and a reputed standing in the community. Now moving on to the social behaviour of the users of SO, they have been divided into group regular, shooting stars and guest and their involvement has been statistically quantified along with the performance of the site compared to the other. The authors have also mentioned alternative solution to the question, but seems like SO has the fastest response time and acceptance time, with valid answer given top priority by voting. The discussion to stop discussion and conversion, plus giving moderator control are the other defining features. We learn a lot about design practices that can help us in the field of social computing, building on the Aardvark implementation, maybe we can create a combination of it with SO to create a stable platform that may not require continuous monitoring. Consider maintain a list online registered users according to reputation and knowledge domain and directing questions to them personally at the same when you post the question to further increase speeds.

Longhao Li 3:10:36 11/13/2014

Critique for Predicting Tie Strength With Social Media In general, this paper talked about how to determine the tie strength by using social media. It is a great achievement in the research between tie strength and social media. I think this paper is important. It gave us the ability to determine tie strength by using social media. Strength of tie includes the amount of time, the emotional intensity, the intimacy, and the reciprocal services. They determined the dimension of tie strength. These dimensions can determine the tie strength. In the paper, the author also did the user study. The result shows that the approach has a very great performance. Determining the tie strength can bring social computing a great improvement. By knowing the relationship between people, researchers, who are working on social computing, can have a great chance to make great social software and give user a great using experience, like find out people who are your friends on the SNS website that are not your friends, or even help you to find people that you may want to talk with. It may connect people more in the society. Critique for Design Lessons from the Fastest Q&A Site in the West. This paper talked about analysis why Stack Overflow has very great success. They also find out how to make website success. In this paper, author mainly analysis why Stack Overflow success. Q&A system performance of this website is extremely high. There are a lot of reasons that lead to this success. The first reason is that there are a lot of user for that website so that questions will have a great chance to be viewed by others. Also the website have a great visibility so that users can easily see the questions and bring the answers. I have a lot of experience of using Stack Overflow. You can easily find the questions that you may want to ask on it so that you can find the answer quickly. But it limited to some common questions. If you always find answer from the website. You always trust the website that they can bring the answers so that when you find that the answer is not in the website, you may want to ask the question on the website. It bring a lot of user to the website, which make the website better.

Yanbing 3:18:47 11/13/2014

The first paper analyzes Stack Overflow, a Q&A site that dramatically improve on the utility and performance of Q&A systems for technical domain. The authors combine statistical data analysis with user interviews to better understand the success. Interestingly, participants in user interviews have founders, site designer so learned lessons are valuable. The research methodology of this paper is kind of unusual to me as the author do data analysis and user interview to understand success of an existing system. However, I like the paper because of two things. First, it brings a detailed data analysis to prove the success of Stack Overflow. The success are measured in many dimensions, high rate of answered questions, fast time of first answer and accepted answer, frequent users post more answers than questions. Those measure are placed in comparison to other QA sites which make the conclusion is very illustrative. Also, it is very important to note a particular characteristic of SO is that it got the critical mass of dedicated users even before the site was introduced. That is really amazing for any start-up system. The founders were able to easily achieve critical mass because both were extremely active in the community and wrote successful blogs. Both has followings that could easily be transferred over to their new project, SO. Also, because the community already existed prior to SO, there was great trust between users, thus resulting in better quality questions and answers. The final factor is their evolutionary approach to design. Instead of sticking to a short bursts of user feedback, as suggested by the iterative design cycle, the designers are in constant communication with the community and pushing updates daily. They designers also used the production site to experiment with new designs, also like SO is a perpetual beta to be improved in accordance with the communities wishes. The second paper aims at using facebook data to determine the ties between two friends on facebook. The hypothesis is that, since facebook treats each friend link as equivalent, one can derive the strength of the friendship from the interconnected facebook activity. This data comes from the interactions of the friends. The that a tie can correspond to the activity on facebook seems intuitive. The more I like someone statuses, write on their wall, and chat with them, the more likely I am a close friend with them. Stronger ties are therefore directly related to the amount of interaction between two friends. Having this information can prove to be invaluable in determining who can see what and have access to what information. Google+ took this idea and made it into circles. They give multiple levels of privacy based on what circle someone is in. Family is allowed to see most everything, friends are similar, acquaintances get a little less, and so on. The models used to predict tie strength show a high level of correlation among close friends in a social network. I do believe Facebook uses data collected from posts, pictures, and other interactive activity between you and friends to determine those that you have close ties with in your network. This kind of information can be used to help improve privacy levels on social networking sites and where sharing information with others. Like the authors mentioned, more behind the screens data would be useful like who friended who would be helpful to in improving the accuracy of personal ties.

Yingjie Tang 3:28:50 11/13/2014

“Predicting Tie Strength With Social Media” is a literature which bridges the gap between the theory and practice of using the social media to accurately predict the tie strength. This is a very novel research before 5 years because the social media has been grown into a very powerful tool which has greatly influenced our daily life. Mark Granovetter introduced the tie strength in his paper “The Strength of Weak Ties”. I have been using social media applications for more than 5 years and I can strongly feel it’s inner power with ties. The people I contact most and I see most of their status will influence me. When I first realize the aim of the paper, I thought about the main work is how to define the tie and quantify it. It is hard to quantify it because their are so many factors that can influence the ties. To solve this problem, this paper first using the original idea from Granovetter and to adopt four factors: amount of time, intimacy, intensity and reciprocal service. However, it is not enough in the social media environment. Maybe we can say that we can take the convenience of social media to create some new dimensions. At pastl, there are a lot of limitations to do user study because the users had to recall their friendship. However, with the social media, we can easily analyze the data from the history of the relationship of the users, and we can take advantage of long friend lists and rich interaction histories. Thus the paper came to 74 Facebook variables as potential predictors variables and the author took the advantages of Facebook’s breadth while selecting variables that could carry over to other social media. After determining the variables, the next topic should be fix their weight to the whole tie value. To address this problem, the paper modeled tie strength as a linear combination of predictive variables because the linear model can help to take advantage of the full dataset and explain the results once it is built. And the research also come across the common obstacle for ego-centric designs that the observations within a participants were not independent.———————————————————————————————————— “Design Lessons from the Fastest Q&A Site in the West” is a literature which analyze the data of Stack Overflow and the interview for the founders to find out the reason why Stack Overflow is so successful. The success of Stack Overflow is great because the result turned out that most of the questions were answered correctly(90 percent) and most of the questions were answered in 11 minutes and the registers were over 110k. I love this paper because in some sense it is more practical in the industry part. Thus I treat it as an inference and a brochure and as if I am going to establish a Q&A forum. The factors for its success will not go beyond the following three: 1)Making competition productive, the tight focus on technical answers enabled by the Q&A format and a voting system created a strong alternative to the existing software forum. One key for the success or the Stack Overflow is that the founders themselves are active software developers who know the technique questions and their format. 2)Credibility in the community. An obvious question to Q&A forum design is credibility and reliable of the answers. One possible solution is to make the registers gain more credits on the credibility which means their answers will be more reliable in the future. And also, mentioned by Prof. Wang, that it will be more meaningful to get the authentication from the social media like Facebook because the answerers will be labeled expert if he answered more correct questions and it can also prevent some malicious fault answers. 3)Evolutionary approach to design. The feedback is very important. What can motivate the users to answer the questions still remain a critical question in the Q&A forum design. What the Stack Overflow did was to give some reward and prioritize features to the answerers after a few iterative designs. Also there exists 5 class questions that the Stack Overflow cannot answer. Those questions on one hand lack the motivation for the users to answer if the questions are tedious to answer or have some obstacles for the users to answer like some obscure technologies for which there are few users on the other hand.

zhong zhuang 4:49:37 11/13/2014

In this paper, the author tries to measure strong or weak tie relationships between users in social media. The paper uses many mathematical methods and tries to prove that these social relationships can be measured. Author recruited 35 participants and let them rate their facebook friends in 4 type of variables, and then use the result to feed mathematical models. The work basically addresses the fundamental challenge for understanding users of social media systems. How do users relate to one another in these spaces.

Brandon Jennings 6:06:36 11/13/2014

Predicting Tie Strength This paper is about enhancing the user experience of social media by accommodating friendships between close friends and enemies. It bring up a point about social media treating relationships as a like versus dislike dynamic when in fact there is much in between. This will become important as social media becomes more of the norm in society. Many people have a few select close friends, then larger group of associates and acquatinences and then the rest of the general public. Most social networking tools assume anyone labeled a "friend" is someone who should have access to your information, which may not necessarily be the case. Questionnaires like the one proposed in this paper will make social media more adaptive to social trends and allow more flexibility for user privacy. Design Lessons This paper seeks to investigate the success of questions and answer forums, specifically Stack Overflow. The high rate of responses and answers to technical questions is important because this can serve as a model for other question and answer forums. Forums like Stack Overflow are important because they allow the community to share information and provide insight into problems and solutions. It also allows people with the same problems and issues to receive information on solutions, as many times people have the same problems. Stack Overflow offers unique features such as letting the community decide which responses and answers are valid and useful, and indicating them as such. This relieves a person of having to sift through a host of responses that aren't useful.

yeq1 7:49:43 11/13/2014

Yechen Qiao Review for 11/13/2014 Predicting Tie Strength with Social Media In this paper, the authors had borrowed the concept of “tie strength” from social science and invented a predictive model that uses characteristics of social media data to make inference on the tie strength. Facebook has been chosen as the primary recruiting site, and both online and offline interactions were used to make predictions and gather private data that may be useful in determining the true tie strength. The analysis is in two parts. Statistical methods were used to capture how useful are different dimensions of tie strength, and how the resulting predictive model perform on the data collected. The authors had found 15 different predictive variables with high betas, the most relevant being the time since the last communication. The prediction model also seems to work well on predicting the ties, with some bias towards weak ties. Furthermore, qualitative analysis on interviews were performed on participant’s responses to determine why the predictive models may err. The authors had determined that asymmetric relationship as well as confounding medium may be the cause. This paper is interesting in several aspects. The first is how to borrow a concept from another field and use it to make contributions. We have seen another example of such practice before. The potential use of providing a system that automatically infers the relationship intrigues me. In fact, this paper, along with another paper that correlates tie strength with access rules, were both parts of my reading two semesters ago. Another interesting aspect is how the authors used qualitative analysis on interview data to reason why the model may not work well all the time. Often, computer science publications simply treats these errors in predictive models as “noise”, with characteristics of the noise left for “future work”. This paper captures that not only we have these kinds of noise, but the characteristics of these noise tells us it is not possible for us to eliminate these confounding factors by simply collecting the data within the social network. The last interesting aspect of the paper is how the researchers can continue to make contributions by providing “workarounds” of the Facebook’s research and privacy policies. Though I think currently the research is too early to be adopted, I think that follow up research will likely to make contributions in making configurations easier to perform, thus allowing the users use some features without teaching them how to learn these features. This is how things should be if we take in the premise of the ubiquitous computing. Design Lessons from the Fastest Q&A Site in the West In this paper, the authors analyzed the website Stack Overflow. They first argued that the current activity of Stack Overflow suggests that the website is very useful. The authors made the argument by collecting user activities over two years, and they had found that a large user base, quick answers, multiple answers, and the active community were some of the good parts of the current Stack Overflow. This argument is used so that the paper’s contribution can be made meaningful to the readers. Furthermore, the paper suggests why the site had gained traction by using a mixture of statistical analysis and interviews (this is yet another time we see this type of paper). The paper argued the main reason is that the developers had actively engaged with the users even before the site had launched, the unique Q&A structure allows askers to quickly read through the answers and vote, and the votes and other related incentives encourage the answers to provide clear, accurate, and quick answers. The founders engage with the users often, and in early stage of the site, daily improvements were made taking into account of issues. The authors, however, do not think that this site is one that is better than all others for social Q&A. They argued that there are some questions cannot be answered as well using this approach, and previous attempts of licensing the site had failed. This suggests that different structures of answers and incentives may be suitable for other types of social Q&A, thus providing future research opportunities.

changsheng liu 8:41:29 11/13/2014

<predicting tie strength with social media> discusses how the closeness of relationships can be predicted. In social media relationships are very binary. Either you are or are not a friend of someone. In real life, relationships exist on a spectrum. The goal of this research was to try and predict where on this spectrum a relationship lies based on the information on the social media site. The technical term for this strength is called tie strength. This measurement has four dimensions which are amount of time, intimacy, intensity, and reciprocal services. To collect data they asked questions of users friends. In addition to these responses, they also collected many variables regarding the users account. Using these variables the researchers tried to predict the actual strength of the relationship as indicated by the participants. Overall, I found the paper to be an interesting look at social media and how it models real social interactions. <Design Lessons from the fastest Q&A site in the west> discusses how the website stackoverflow became such a useful resource. When stackoverflow is compared to many similar sites, we find that it provides better answers within a shorter period of time. The authors are curious to why this is, and if the same techniques can be applied to other sites to improve there performance. Three things that really contributed to the success of stackoverflow were that it leveraged competition, formed a sense of trust within the community, and focused on evolving the design. The authors discuss the nature of stackoverflow, and its performance. One thing they noticed is that the answers have always been fast, which suggests that they obtained a critical mass of users even before they started. The authors grouped users of stackoverflow into four categories. As it turns out, stackoverflow is a niche Q&A site in that some questions are not suited for it, such as ones discussing obscure technologies, ones that are tedious to answer, and those which do not have a single best answer. In addition to the main site, there is a meta site to discuss the direction of the main site. The authors sugest that this helped keep the development cycle short, by carefully listening to its users. By taking this style of design, the authors hope to be able to improve similar sites. Overall, I found this paper interesting because it looked at an interface which was successful, tried to figure out why, and then suggested how these ideas could be used to improve other sites.

Christopher Thomas 8:44:00 11/13/2014

2-3 Sentence Summary of Predicting Tie Strength with Social Media – This paper presents a model for predicting a tie strength with friends on Facebook. Their predictive model uses a variety of features, including checking for intimacy words, intensity of communication, duration, etc. Then, they present some use cases for the model. I found this paper very interesting. The authors found that by using a variety of predictive simple predictive variables, such as educational differences, use of intimacy words, the strength of mutual friends, social distances, etc. they were able to automatically determine the “tie strength” between two users on Facebook with their automated model. This research has numerous implications and I believe Facebook has been integrating similar approaches to this now. This is research that has practical applications as well. For instance, I know that Facebook now allows you to contact “close friends” if you need help getting into your account. If a user is locked out of his or her facebook and is unable to regain access or even access their email to reset their password, Facebook can determine who the users close friends are and send a request to verify. The close friends will then call the person and make sure that they are the ones trying to reset their password. When a few close friends approve, the person is able to login again and reset their password. This research also has national security implications. If someone corresponds regularly with a known terrorist, automated systems could determine the closest ties to the criminal. Those people could then themselves come under investigation. Some limitations on this research are that it does not have access to the data Facebook itself has. The model would probably be much more predictive if it were able to determine things like which friend added which friend. Still, just using publically available data, to receive a 88% performance is pretty good.   2-3 Sentence Summary of Design lessons from the fastest q&a site in the west – This paper analyzes the StackOverflow website, where users can pose programming comments and questions. The authors argue that Stackoverflow model works particularly well. The authors explain why they believe this and explain how other companies can benefit. The authors used statistical tests to analyze StackOverflow and tried to explain why it does so well. My personal experience with the site is that most questions get answered very quickly. In fact, I don’t believe I have ever asked a question on the site that wasn’t answered in a few minutes. This does beg the question, how is that happening and why isn’t it happening on other sites? To explain this, the authors argue that StackOverflow’s game model and reputation system help explain it. We also talked about last lecture that sometimes you release a product to a closed community. Professor Wang mentioned how Quora started only with a few people and in this paper, the authors show how credibility permitted StackOverflow to gain dedicated users and experienced question answerers. The authors argue that this may be a pivotal point behind it’s success. Thus, we can see StackOverflow as an exercise in social computing. This shows that while HCI research is useful, companies also develop innovative HCI paradigms on their own. Personally, I feel that the key behind SO’s success is the reputation system. Also, I believe a key component is the reward system for users. I notice similar behavior on the Matlab forum. It would be interesting to extend this research to new domains and see if the same behavior is observed across domains, or if this is something relatively isolated to the tech community, who may care more about internet reputation than others.

Jose Michael Joseph 8:52:08 11/13/2014

Predicting tie strength with social media This paper is about predicting the level of friendship between two people based on their interaction on Facebook, a popular social media site. It uses the concept of strong and weak ties to relatively judge the friendship between two people. Loose acquaintances are known as weak ties whereas trusted friends and family are known as strong ties. The dimensions used to define tie strength vary significantly. In theory though tie strength has at least seven dimensions and many manifestations. The method of data collection relied on querying users for information by using a GreeseMonkey plugin which is a Firefox web browser plug in. Users rated their friends on various categories and this information was then used to understand how close they are. While the user entered data about their friendship, the author’s system automatically collected some data based on predictive variables, intimacy variables, duration variables, reciprocal services variables, structural variables, emotional support variables and social distance variables. It then analyzed these methods using statistical methods which measured the various parameters that the system automatically tested for. This value would give the predicted friendship level between the two individuals. Personally I feel this system has many flaws. It is hard to determine the friendship between two people based on the context of a single social media platform. I personally talk to all my close friends off facebook and thus such practices would skew the analysis of the data received. To compensate for this a system should be made such that it also searches other complementary social services such as messages and IM. This would help the system better quantify the friendship between two people and have a much higher accuracy.

Jose Michael Joseph 8:52:37 11/13/2014

Design lessons from the Fastest Q&A Site in the West This paper talks about the various design reasons that could have led to the success of the Q&A site Stack Overflow and which resulted in questions being answered in a median time of 11 minutes. The authors of this paper have realized that the fastest answer times and high answer quality arose from the efficient use of a reputation system and a strict set of community guidelines that favor factual answers. These decisions were the consequence of a design philosophy that was set by the founders. Three factors that the authors believe are critical to the success of SO are making competition productive, credibility in the community and evolutionary approach to design. SO focuses on having few concise answers as opposed to large number of partial answers. Thus the average number of answers per question on stack overflow is lower than other traditional Q&A websites such as Yahoo! Answers. But each answer can also have comments and this leads to measuring another value known as the thread length. The thread length of stack overflow is equivalent to that is Yahoo! Answers and other such sites. The authors also noted that another popular website. Aardvark, is faster than stack overflow in generating answers. This is because this website routes the questions to those users who are already online. Thus the median time for Aardvark is only 6 minutes and 37 seconds which is close to half of the median time for Stack Overflow. But seeing that this is a simple system to implement, one wonders why Stack Overflow has not done the same. The authors also state that the response time of answers has always been fast and thus there is very little room for improvement. This is backed by the data that the answering time remained more or less the same irrespective of the large number of users that have since joined Stack Overflow compared to when it was in its early days. The authors also classified users into different categories which are community activists, shooting stars, low profile users and lurkers and visitors. They have also stated that while low profile users can be successfully converted into being an active part of the community, SO has still been unable to make lurkers and visitors into an active part of their community and that most of these users move on after an initial phase of infatuation. The authors also praise the high level of design iteration that takes place in SO. The site is almost republished every day with minute changes. The developers of the site continuously look for feedback from its users and implement those on a priority basis depending on the number of votes assigned to each feature. SO still has many challenges it can overcome to streamline the overall user experience and thus to attract more users. One such common problem for the users is that it is often not worth the effort to correct an already existing users as the “pay out” of such a task is not as great as that in creating a new answer. Thus often partial answers are left uncorrected on SO. This is one problem that definitely has to be addressed to maintain the standard of the website.

Xiyao Yin 8:54:03 11/13/2014

‘Predicting Tie Strength With Social Media’ mainly discusses a topic social science which has investigated for decades under the theme of tie strength. Authors’ work bridges this gap between theory and practice and presents a predictive model that maps social media data to tie strength. They first review the principles behind tie strength and discuss its proposed dimensions. Next they present the construction of their tie strength model, after careful evaluation, they discuss on some limitations in their ideas. The strength of a tie is a (probably linear) combination of the amount of time, the emotional intensity, the intimacy(mutual confiding), and the reciprocal services which characterize the tie and there are four tie strength dimensions including amount of time, intimacy, intensity and reciprocal services. Based on these ideas, authors start their experiments and evaluation. They use examples from Facebook and compare their prediction with users’ true concern. Results show that social media can predict tie strength. In my opinion, this paper does a good job in evaluation but still has some limitation. Since this work looks only at one social media site, more experiments on other sites need to be checked in the future. ‘Design Lessons from the Fastest Q&A Site in the West’ analyzes efficiency on Stack Overflow,which is a privately held website, the flagship site of the Stack Exchange Network, created in 2008 by Jeff Atwood and Joel Spolsky, as a more open alternative to earlier Q&A sites such as Experts Exchange. The name for the website was chosen by voting in April 2008 by readers of Coding Horror, Atwood's popular programming blog. Results show that Stack Overflow has quite a huge success and authors find that high visibility and daily involvement of the design team within the community they serve as well as a prior superior technical design are quite significant reason. In this paper, author first conducted a statistical data analysis of the entire SO corpus to understand usage patterns. After several research, they found there factors which are critical to the success of SO: Making competition productive, Credibility in the community, Evolutionary approach to design. In conclusion, this paper successfully proves authors’ assumption for the site’s success. As far as I am concerned, this paper is not just a good research but also a good guide for me. Since I am still new in the US, I don’t have much experience in using SO and I haven't realized such success in this website. I can start to use it now.

Vivek Punjabi 9:39:11 11/13/2014

Predicting Tie Strength With Social Media: In the past decade, social science has been investigated and studied under the theme of tie strength. This paper brings it into practice by presenting a predictive model that maps social media data to tie strength. They try to predict tie strengths for over 2000 different social friendships/ties asking some questions about certain friends. They have presented 32 of 74 potential predictors of tie strength from Facebook. The questions were the dependent variables that were answered by the users. In the results, their model performs best for 2 questions of $100 Loan and how strong. They also conducted interviews with users for which predictions didn't match. In the end, they found some obvious and some surprising predictive variables, along with some inherent exceptions, that affects tie strengths. This paper creates roots for many new ideas towards social science that help understanding the users of socio-technical systems. Its surely an interesting topic with lot of opportunities in research of social computing and data gathering. Design Lessons from the Fastest Q&A Site in the West: The paper analyses the fastest Q&A site of the west , Stack Overflow. They use statistical data analysis and user interviews to understand their success. They used the publicly available SO database export files and found that certain features like fast answer times and high answer quality were critical to its effective functioning. The three factors that were found critical to the success of SO are making competition productive, credibility in the community and evolutionary approach to design. Some interesting statistics about how SO works ere found such as answers in 11 minutes and frequent users post more answers than questions. There are some question types not well supported by SO. There have been a small, active base in the meta site as well. The site face some challenges as well which may still need modification of the design model. Also, community organization matters more than design. At the end, the author arises a question whether systems research can play a huge role in social computing. It in indeed a motivating paper as SO is one of the most used resources by computer scientists. Various factors identified in this paper can help to think in an alternative and more creative way for developing Q&A sites for any knowledge domains.