Assessment 3: Group Portfolio
Do intervention videos have an impact on reducing gender bias in STEM fields?
Please click the button below to visit our website:
WEEK2:Website Building
This week I learned the basic components of a website,
and tried to build my own personal website using code.
It was hard but rewarding for me.
1.what i have learnt/done:
·Understanding of html and css
·Familiar with basic html language
·Trying to build my own personal website
2.what i found is difficult:
·Proficiency in html and css
·basic website architecture
·website style and artwork requirements
3.feeling and reflection：
First of all, during the learning process, I realized that I don't know much about websites. I could easily use it in my daily life, but had not looked at it closely. It turns out that there is a lot of clever design and complex code behind a simple looking website.
When I first started to learn how to write code through online courses and workshops, it felt quite easy. Isn't it just a matter of memorizing various code elements and using them skillfully? When I started to make a website myself, I realized that I need to consider different factors, such as website functionality, basic structure, operability, and aesthetics. Each of them is not easy to realize.
Secondly, it was my first time to learn a programming language and it was very difficult. From the beginning of recognizing the code of the existing website, to installing the software, configuring the environment, and actually starting to knock out the code, every step of the way, all kinds of strange errors will appear, it is very time-consuming, but do a good job to see their own website to present the content they want at the moment really super sense of accomplishment.
Also, I've found a lot of tutorials on html and css on the web. When I have an effect I want to achieve but don't know how to, Google it and there's always someone sharing his code in various forms that I can imitate and learn from.
Overall, so far I'm still slowly figuring out how to learn, and I hope that after one semester, I can build a good-looking and useful website by myself.
WEEK3:Data Scraping
Learning data crawling is a long process,
and I should still be in the introductory stage at this point.
For this study I tried using Scrapehero and Octoparse as tools to crawl the playback data of every video on my personal social media platform dounyin.
1.what i have learnt/done:
·Learn about different data scraping software
·Experimenting with different software to capture data
2.what i found is difficult:
·Different data capture software has different modus operandi and limitations and requires constant experimentation
·Some data cannot be captured due to privacy settings.
·The captured data is usually messy and needs to be cleaned and formatted.
3.feeling and reflection:
This is my second time learning about data capture. In order to conduct quantitative research at undergraduate level and provide data support for my research, I had learned to use Octopus to crawl the data required for my research. But there was no deep learning.
This time, I learned the html structure of the web page in advance, and I obviously feel that it is much easier to use the software for data crawling, I am more clear about the code structure of the web page, and I can locate the data that needs to be crawled for crawling more quickly.
Also tried using Google CoLab, which was recommended by my teacher, but I don't know why I was dropping out and disconnecting so often. I need to look into it a bit more. And I'm not too familiar with python, so it's still a bit difficult to use. But he's super convenient without having to install it, and it's a pain every time I install the software, configure the environment, and run into version compatibility issues.
WEEK4:DATA ANALYSIS
This week has been all about exploring data and data analysis, and I have a deeper understanding of the power of data and data analysis.
Counting is a privilege.
In last week’s workshop, our group chose to study the topic “The Impact of Digital Engagement of Colleges and Universities on Employment Rates” and collected data such as the one shown above mainly from official reports and university social media.
Reflections are as follows:
1. The data we collected are all from public platforms, and do not involve privacy and ethical issues.
2. There is a subjective gap in our data source. Our supervisor assumed that colleges and universities would choose Instagram as the main channel for publishing employment-related information. Perhaps different colleges and universities have different social media.
3. Some international students may not use Instagram and are not counted by official data. Our data statistics may make this group invisible and ignore their interests. Feminist geographer Oni Seage said, "What is counted is important." Data often becomes the basis for policy making and resource allocation. Ignoring this part of international students will prevent them from being considered in the school's subsequent employment service improvement plan.
In a 2018 essay, “Design Justice, A.I., and Escape from the Matrix of Domination,” they give a concrete example of why design justice is needed in relation to data. Therefore, in the process of data collection, classification, and analysis, we need to fully consider the fairness of the data, whether there are ethical issues, whether there is subjectivity, and whether some groups are ignored.
WEEK5:DATA VISUALISATION
This week I have been learning about what data visualisation is, why it is important and trying to do it myself.
In fact, I have learnt data visualisation at my undergraduate level, but I used to think that data visualisation is just a simple way to show complex data in charts and graphs, so that the audience can understand the content of my data more easily and more convincingly. But through this week's study and after reading Helen's paper, I found that I have overlooked many factors of data visualisation.
What is data visualisation?
1. In essence, data visualisation is also a right. Depending on the purpose of the visualisation operator, parts of the data can be made prominent or made invisible. Data visualisations are not neutral windows on data: they prioritise certain points of view, perpetuate existing power relations and create new ones, they do ideological work. (Helen, 2016)
2. Data visualisation as a visual tool can be used to simplify data, but also to present data in a more vivid and nuanced way and to express positions and emotions. For example, in The Telegraph's 2018 interactive feature ‘Born Equal. Treated Unequally’, the gender gap in the UK is examined in a number of dimensions. Pink and blue are alternative hierarchies, and members of the Telegraph team aimed to mitigate inequality rather than exacerbate it, so instead of using the traditional blue-pink to represent gender, they used colour to challenge stereotypical male/female colour coding. Visualisation is often seen as a way of reducing complexity, but here it had the opposite effect - taking simple, depressing ideas and making them more complex, nuanced and just. (Catherine, Lauren, 2020)
3. Used well it can enhance the power of data.
How to do data visualisation?
1. Consider the audience. The purpose of a visualisation influences how it is put together, and the style and tone of the visualisation should vary according to the expected level of audience interaction and engagement (Helen, 2016).Kirk (2016) argues that visualisers need to consider whether the visualisation will explain key insights, enable users to interact and explore, thus finding their own insights, or exhibit the data visually with users left to do their own interpretation (he defines these as explanatory, exploratory and exhibitory purposes)
2. Consider chart type, colour, annotation, interactivity.
WEEK7:MACHINE LEARNING AND FACIAL RECOGNITION
This week we focused on understanding and learning about machine learning and facial recognition. And exploring Machine Learning using Teachable Machine in Worship.
Each of us knows that our daily lives are actually filled with machine learning and all kinds of automated facial recognition, to the point where it's become so routine that I've ignored them. This is the first time I have taken a serious look to understand them.
Facial recognition technology, despite its widespread use in many fields, has also revealed a clear bias, and the contemporary automated essentialisation of sex/gender through faces is both racialised and trans-exclusionary: it asserts that the masculine/feminine binary is biological and fixed, and elevates the white face as the ultimate model of sex/gender difference (Morgan, Madeleine, Alex 2021)
1. Racial bias. Facial recognition algorithms perform differently in identifying different races, especially in identifying people with darker skin colours with higher error rates.
2. Gender bias. Facial recognition algorithms often show inaccurate recognition of women, especially when it comes to women with darker skin colours. In addition, non-binary or transgender people face greater challenges because many systems are designed based on traditional binary gender models that ignore the presence of diverse gender expressions (Buolamwini and Gebru, 2018; Raji and Buolamwini,2019).
3. age bias. Facial recognition techniques are less accurate in recognising older people and children because there are usually insufficient samples of these age groups in the training data.
4, Cultural bias. Some facial recognition systems may be insensitive to certain culturally specific facial features or expressions (e.g., makeup, headscarves), leading to recognition errors or bias.
5. Bias in the context of use. Facial recognition technologies may exacerbate discrimination in certain scenarios. For example, the overuse of facial recognition technology in surveillance systems in some cities, particularly in minority communities, can lead to excessive law enforcement and biased behaviours, as exemplified by the Chinese government's facial classification of Uyghurs into ‘re-education’ camps (Mozur, 2019).
WEEK8:ALGORITHMIC IDENTITY AND REPRESENTATION
This week explored the relationship between digital media and the body. In the lectures examined how digital media affects our expression and interaction. By looking up our own INPUT, OUTPUT and PROCESS in social media in the workshop, we are continuing to explore how algorithmic identities are formed and reflecting on whether algorithmic identities can be representative of real identities.
The reflection of reading
Cheney-Lippold, J. 2017. Introduction. In: We Are Data : Algorithms and The Making of Our Digital Selves. New York: NYU Press, pp. 1-32
1. what is algorithmic identity?
Our algorithmic identity is realised through data, and only data. It is a process of collecting from a database; it treats our behaviour as data, our social relationships as data, and our bodies as data.
2. Characteristics of an algorithmic identity?
·Fluid: as we are constantly producing data, our algorithmic identity is constantly changing.
·Diverse: Hundreds of companies assign different ‘genders’, ‘races’, and even ‘classes’ to each of us based on different algorithms, so we don't have a single, static sense of ourselves, but rather a myriad of competing, regulating identities. Instead, we have a myriad of competing, moderating interpretations of the data that make up who we are.
·Inauthentic: the data of who we are or who we might choose to be is more important than who we really are.
·Undetectable: In the ‘one-way mirror’ metaphor, most internet users remain unaware of how their data is being used, while website owners enjoy near-universal access to that data. Our algorithmic identities are similarly formed behind a one-way mirror: it is largely impossible to know what our ‘gender’ is, how ‘old’ we are, or whether we are ‘at risk. ’
·Profitable
·Not subject to our control
Sumpter, D. 2018. Chapter 3: The Principal Components of Friendship. in: Outnumbered: From Facebook and Google to Fake News and Filter-Bubbles - The Algorithms That Control Our Lives. London: Bloomsbury
1. How does Sumpter (2017) classify his friends?
Sumpter categorises his friends into the following three main categories based on their Facebook posting habits:
·Focus on personal: friends who mainly post about private life content, such as family, mates and lifestyle activities.
·Work-focused: Friends who primarily share work-related content, including career updates and work activities.
·Focus on social: Friends who focus on public topics such as news, politics, culture and broader social issues.
2. Do you think he is missing any categories?
Some of his friends don't like to post and he can't analyse them
Three simple categories don't tell the whole story
Each friend has different Facebook habits (e.g., work apps, life-sharing apps), and Facebook identities are only partially representative of the friend's identity, not the real personality of the real friend.
my thoughts and questions
1.Does algorithmic identity affect our perception of our self-identity?
2.Can I actively change my algorithmic identity?
3.Are people changing their Internet expression based on algorithms?
WEEK9:IDENTITY AND REPRESENTATION
This week we researched and discussed Sumpter's Method and online communities（Still updating, I am re-experimenting this method）
Reflecting on Sumpter's Method
In last week's seminar, we shared our new findings and limitations in using Sumpter's Method to classify friends' identities.
The common points we found are:
1. This classification can indeed reflect part of a friend's personality and identity. From the analysis of friends' social media, we can simply see whether they like to share and what they like to share, which partially reflects their online or offline identity. It can be speculated that algorithms can collect more of our behaviors and data and make more detailed classifications, which can more accurately reflect our identities and thus conduct different commercial and political behaviors.
2. However, this method cannot fully reflect. This is because friends have different habits of using social media. Some people don't like to post, and some people will deliberately set their own social media personalities and thus continuously post on specific topics (such as beauty bloggers, pet bloggers, and self-discipline bloggers), so it is difficult to analyze the true identity of friends through online data. Similarly, we deceive the algorithm or deliberately disrupt the algorithm, just like Google Ads' speculation on our identity is not completely accurate.
3. This method is very difficult to operate: 15 categories make it difficult to subdivide all posts, and the text, pictures, and videos of posts sometimes do not express the same meaning, so the classification is very subjective, and the algorithm also categorizes us according to its needs.
Extending Sumpter's Method
1, Referring to Sumpter's (2018) methodology, I briefly categorised and plotted 2D coordinates of 15 recent posts by my 32 friends. Two of these categories were then plotted against each other.
From the 2D coordinate graph, I can briefly see the identity characteristics of my friends. One of the friends likes to post content related to celebrities, another likes to post about her family's pets. Some other friends like to post content about food and sports related to their lifestyle habits. Some of them are keen on posting the same topic type and some of them are very scattered in their post topics.
It can also be noticed from the graph that there are interrelationships between certain categories. Those who like to share about daily life will share more outdoor-related content, but not usually news and reflections. People who like to share the typology Daily Life and Outdoor will also post more adverts.
2、I classified the 15 posts into two categories: personal and others (as shown in the figure below, white is posts related to the individual, and blue is posts related to others, nature, and society).
I discussed this finding with some of my friends. One friend, whose posts were almost all about herself, admitted that she often shared her personal life and searched for personal information on this social media, but because this platform is a life sharing platform, she would share other content on other social platforms, so she did not think she was a self-centered person. This also proves the limitations of inferring identity based on data from one platform.
3. in-depth interviews with 5 of my friends
What information do you think I will find out about your social media identity?
Do you think there are any anomalies in my results?
Were you surprised by these results? Why?
Would you change your behaviour on social media afterwards because of these results?
Do you think this data is representative of who you are?
4、I think there are several ways to expand this method to make up for the limitations:
Multi-platform cross-research.
Collect more interactive behavior data (likes, comments, and reposts)
Ethnographies of Online 'Communities' (combined with reading)
1. Online community members
In Rob Kozinets' (2010) "Netography" method, he suggested using the "continuum of participation" to define "community membership", including "self-identification as a member, repeated contact, mutual familiarity, shared knowledge of certain rituals and customs, a sense of obligation and participation". This is consistent with what we discussed in the seminar. For example, a classmate is a member of a fan club of a certain star, and the fan club is an Online 'Communities'. They have their own fan club name, like the same star, have their own support activities, and communicate frequently on a certain platform, so they recognize each other. They even have obvious distinctions between community members, such as loyal fans, passerby fans, etc.
2. Online communities have network sociality, which is fast-paced, fleeting, and cross-regional in nature (Andreas Wittel, 2001).
3. Online communities are threaded social (Postill and Pink, 2012). They provide people with a broader platform for expression and give them a stronger sense of belonging, but they may also lead to more intense verbal conflicts and have a great social impact, but they are also manipulated by algorithms.
