Images, Art & Video - Collins 2009

[jamesallenevans]

Collins, Randall. 2009. “The Micro-sociology of Violent Confrontations”(Chapter 1) and “Confrontational Tension and Incompetent Violence” (beginning of Chapter 2). From Violence: A Microsociological Theory: 37-43. (see Canvas for reading).

[lkcao]

This piece provides very interesting information about how violence proceeds, and how we produce myths about violence. The emperical evidence comes from observations, records and reconstructions, and pictures contain valuable insights too. My question is: Is 'violence' of different sources really the same thing? People often categorize many different social facts under the same concept. But they are often different in essence. How can we confirm that some social facts are really comparable?

[ckoerner648]

This reading is so rich that it even inspires me beyond the question of images, art and video. It gives me a new idea for text analysis. Collins notes that interviewees “are generally not good observers of the details in context of dramatic events. Our ordinary discourse does not provide the language in which to describe micro interaction well; instead, it offers a set of clichés and myths that predetermine what people will say” (p. 4). I think that it is very likely that this insight holds beyond violence. When talking about work in oral histories, people may also fall into clichés that do not necessarily help us understand how they experience a work situation. I take heart, however, in Collins’ observation that “we become better at [...] asking retrospective observers for the kind of details we would like to know [...]. By providing a vocabulary we make our informants often quite good reporters of details they otherwise gloss over” (p. 6). Long term, I now consider complementing the oral histories collected by others with collecting my own oral histories using a Collins-style vocabulary and then analyzing it with computational content analysis tools.

[tzkli]

The reading is very interesting and eye-opening. The author makes a compelling case for the inadequacy of evolutionary theories of violence. Some of his arguments in this regard, however, seem to confound biological evolution with institutional/social "evolution" (p. 25-29). Evolutionary biologists disagree on the time scale needed for a trait to evolve to be discernibly different, but few use the term as loosely as the author does, for example in his use of the phrase "evolved biological propensity" (p. 26). It is a bit ironic that the author dismisses the Bourdieusian term, "symbolic violence", as "mere theoretical word play" (p. 25) while he himself seems to have made a similar mistake.

My question then is about the divide between positivism and interpretivism. Interpretivism seems to be very helpful in alerting us to the important intricacies and complexities of human behavior that the positivist's Occam's razor more often than not erroneously trims away, but how can we assess the validity of interpretivist findings to make sure the theoretical intervention is not "mere theoretical word play"?

[katykoenig]

The author's theory of violence as a situational process was interesting and seems in line with various public policy theories regarding the role of institutions and systems on people's decisions. How could such theories of social interactions be incorporated into computational approaches, e.g. NLP and ML, to address policy issues or understand societal dynamics?

[rkcatipon]

The author's theory that violence is a situational process is fascinating. I agreed with the claim that most violence with intent to harm is actually rare and a process

One of the aspects that I wish the author delved more into was the mythologies around fighting, which was mentioned in regards to the group brawls that sometimes spark after sporting matches. I think there is potential to study violence as a mechanism for acculturation. For example, I used to practice a form of traditional Filipino knife fighting called Escrima or Arnis. In practice, we had the intent to hurt each other but it was within the context of "play"-- which the author posits as a different form of violence (such as the pillow or food fight) that encourages wider participation. It is that encouragement that is important to forms like Arnis. It was violence, true, but the martial art was also a cultural practice that communicated histories of occupation and migration. I would be curious to see where the author would place violence as a method of negotiating identity within the examples given.

[luxin-tian]

These two chapters provide me new perspectives from sociology to understand violence by putting the researcher in the shoes of not only an investigator, or a researcher, but also the agents involved. As is can be imagined, the theoretical examination of the myth of violence may be incorporated into the machine learning technology to perform real-world tasks on an unprecedented scale. While violence is some sort of consequence or burst of tensions or conflicts between agents, in the modern real world there might be other forms of expression that are not violent. There can be similar interactions as expressions of tensions or conflicts in a diplomatic context or cases such as non-violent resistance, which can be hardly identified or examined directly using the theory constructed for direct or obvious violence. Furthermore, what might be valuable information is that such forms of tension can sometimes be predictors or prior status of violent interactions. I wonder if it is possible to further extend the scope of the investigation of violence to its more complete lifecycle, and where statistical learning can be applied to gain insights?

[ccsuehara]

According to the author, there are three methods for approaching situational violence: Recordings, reconstructions and the human observation. I would like to know more about the biases each method has, and what is the role of the three of them complementing each other in the interplay of the analysis.

I like the part where the author takes power out from recordings (the surveillance society, cameras, phone tracking, etc) and gives it back to reconstructions and human observations, but has this changed give the shift in technologies?

[laurenjli]

The author mentions that pictures can be more useful than video because they can be used to closely examine emotion. I wonder if video can be used more effectively now given the progress in computer vision and other image recognition methods? Would it be possible to employ a computer vision technique to pick up on certain facial expressions/cues and tie them to emotions to assist in this type of research? The author mentions needing a certain amount of background in the field to be able to be able to pick up on certain emotions and expressions and I wonder if a model could be trained to do so.

[deblnia]

I thought this reading was interesting, and provides a rich theoretical frame for applying computer vision in novel, if troubling, settings. The sort of observations Collins' makes (e.g. on page 27 -- "their postures and facial expressions were crouched and fearful ... even the most active firers are tense and strained", or on pg. 45, "...disfunctioning bowels") are things that we, as both biologically and socially evolved people might immediately recognize as correlating to a facial expression.

But does that mean that any kind of computational technique trying to discern these nuances would need a human-labelled dataset? Or are there other ways to encode these visual signs of discomfort/tension?

[HaoxuanXu]

This paper argues that violence is a very sporadic event, one that occurs actually in very low frequency. This makes similar to extremely imbalanced data. How do NLP and other machine learning methods tackle this imbalanceness without producing biased predictions?

[sunying2018]

This book mentions three methods for getting at situational details of violent interactions: recordings, reconstructions, and observations. "The three kinds of situational evidence fit together. They complement each other not only ethnologically but also substantively. " I am wondering to develop a common theoretical framework, how should this framework incorporate these three methods properly and what determines the combination?

[jsmono]

The author provided a comprehensive analysis of the culture of violence. He also mentioned the importance of context to the application of violence and pointed out that violence is difficult to carry out because of fear. It made me wonder what if violence becomes a norm, such as during revolutions and social movements, will people feel easier to carry out violence?

[chun-hu]

I'm wondering if we can apply computer vision methods to detect emotions in videos and pictures? And in order to do this, do we need to input a human-labeled dataset with facial emotions to train the model?

[bjcliang-uchi]

I really enjoy this reading! I am a bit concerned, however, in terms of the philosophical foundation of this micro-sociology theory, in the sense that it is extremely individualistic. Of course situations matter, but in historical conditions such as Nazi Germany or China's cultural revolution, when people were brainwashed to an extreme level, they do not behave violently out of any rational concerns or fear as we observed in a daily basis/ military battleground.

[heathercchen]

This paper really differs from other oriental readings we have in this quarter. I am wondering about what will be the outcome if we utilize methods that we cover this week to analyze violence. Do images that have violent content contains some specific elements (just as the theoretical construction proposed in this chapter) if we train them in real data using machine learning methods?

[alakira]

The piece reminds us that violence is a process which is usually fall short of explanation. I am wondering how could we enhance the way of describing / analyzing violence by using computational tools and further intervene it with public policy perspective?

[skanthan95]

What are some ways in which we can use computational psychology to track changes in historically sanctioned violence over time? For example, the evolution in seppuku tradition (by context) that the authors describe?

[yaoxishi]

As argued in the piece, violence is a situational process and can be approached in three ways such as recordings, reconstructions, and human observation, I am wondering how computational algorithms can capture the situational components in these materials to identify violence.

[wunicoleshuhui]

This reading has given me a lot of insights on how violence is and could be interpreted and analyzed. I'm quite interested in applying some of insights on analyzing violence in movies. What are some of the ways to arrange programming tools to distinguish violence in PG-13 films from R-rated films based on the tension and psychological shock the violent scenes may bring to the viewers, so that we could have a more comprehensive movie rating system?

[sanittawan]

The part that I really like about the reading is where the author demystifies fights and how it is related to violence (though I was perplexed how he could possibly have left out the fight scene between Nada and Frank in They Live!). However, my question is about the link between micro and macro theories to explain violence. How can a micro-situational theory that the author proposes explain, for example, the act of terrorism? I am aware that this might not be an entirely fair question for the reading since it seems to be focusing on humans as a unit of analysis, but I am curious about the possible link.

[YanjieZhou]

I am wondering that how different it is to detect emotions than to detect which person the face belongs to, which I think is more applicable and I am truly curious about how this technique is applied.

[arun-131293]

This paper mentions violence is a sparse occurrence. I would like to point out this is a function of the spatio-temporal unit of analysis. For instance, if the spatial unit of analysis is large(a large area), but the time unit is also large(one day), violence will become more 'frequent'. How do we determine ideal units of spatial and temporal analysis?

[ziwnchen]

My question for this paper is also on the computational approach/implementation fo the violence theory. Apart from the real-world scenario, I am also wondering how could this approach be implemented to digital content on online/social media platform?

[kdaej]

One can only capture a part of a violent situation, not the whole picture. Also, I think, though it needs to be stated carefully, the situation can look different depending on where you are standing. Consequently, categorizing the kinds of violence can be only superficial reducing the human experiences. Every situation is different and the narrative of the same event can vary by who speaks of it. I wonder how effectively the discussed objective measures of violence can represent the subjective human experience of violence.

[VivianQian19]

The main take-away from looking at violence through a micro-interactional perspective as suggested in the chapters is that violence is not easy for individuals to carry out. Collins argues that antagonistic confrontations turn into violent interactions as a result of conditions that circumvented the barrier of tension and fear, which reorganized the emotions of participants) (chapter 1). While his theory is generalized/extended through sources on armies, he posits this theoretical model works also in other situations such as police violence and even domestic violence; the author writes that these forms of violence follows the pathway of attacking the situational weak (ibid). He writes that this set of violent situations differs from violent situations when the emotional energy is on the audience instead of finding a weak victim such as in sport violence. My main question after reading the chapters is how we can apply Collin’s micro-interactional perspective to a computational analysis of image, video, and audio data? Are there limitations of applying this perspective to such an analysis?

[cindychu]

The analysis of characteristic or situation account of violence is very interesting and corresponds well with a classic debate of psychology, the origin of behavior and how much environment could influence a people. In violent behavior, more evidence in psychology also supports the ‘environment’/‘state’ account, for example, ’Broken Windows Theroy’.

This chapter also creatively and meaningfully integrated modern visual technology in studying violence, kind of like ‘technology ethnography’. I am also wondering, in text situation, like ‘chat forum’, is there anyway to analyze ‘verbal violence/abuse’?

[cytwill]

This is really a long paper, I think it is more of a book. And the description of violence is in multiple aspects and formats. I am wondering if we can use some computational methods to qualitatively describe the different types of violence or related phenomena introduced by the author?