Header image from Towards Data Science
In the beginning of this workshop, we had a small discussion about the terms Artificial Intelligence (AI) and Machine Learning (ML).
More specifically, we were questioned about what those terms mean, and "which is a subset of which?"
I already knew that even though AI and ML are often used interchangeably, ML is a subset of the broader category of AI. However, this conversation was quite interesting for me because even though my previous degree was in Informatics, Machine Learning was an optional course that I hadn't taken so this whole scientific area remains undiscovered for me. I realized that my knowledge around ML and AI is very scattered, and I have a bunch of complicated terms in mind such as recommendation systems, unsupervised vs supervised learning, convolutional neural networks, ML algorithms e.g K-means, but I am not able to connect those terms and make more sense of them besides roughly knowing what they mean.
This workshop's brainstorming session intrigued me into finally looking more into the theory behind ML and "connecting the dots"!
Overall, the workshop was divided in two main themes:
1) Building Machine Learning Apps
2) Exploring Social Media Mechanisms
Below the activities of the workshop are presented:
Mission. Make your own classifier, either independently or in a group of 2-3 people
Your task is to make a functional classifier application, following the steps below. In the last step, the application is returned as a .zip file here in the Howspace environment and distributed via a QR code to other group members for testing.
You can use AI applications to find out what could be classified with an application capable of classifying images and postures and what kind of problem you could formulate in the first step.
Presentation: Prepare a 5-minute presentation that includes: (share ppt/pdf/etc below)
A description of the problem you aimed to solve.
How you collected and curated the data.
The training process and challenges you encountered.
A live demonstration of your app, showing how it recognizes images and responds with actions.
Reflection: After presenting your app, reflect on the following: (discussion below)
What challenges did you face in collecting and curating the images?
How well did your model perform, and what improvements did you make?
What did you learn about the machine learning workflow through this process?
We were a team of three, me, Aaisha and Kristian.
(1) We first began brainstorming for simple problems that we could solve using a classifier. We had ideas such as creating a classifier for interpreting sign language. Aaisha then posed the problem that since she came to Finland from her country, she often had issues in distinguishing vegetables when shopping in the supermarket, because they look different than they do in her country and the labels are all in Finnish. So, a classifier could solve this problem for internationals coming to Finland. We all thought it was a good idea started creating a classifier for vegetables.
(2) We started with the training data. We split vegetables into categories e.g leaf vegetables, so everyone individually searched the internet for different images o. We picked similar-looking vegetables, for example celery and spring onion.
(3) Everyone uploaded their images onto the GenAI trainable machine and gave them appropriate labels.
(4) We then combined our training data into one project and trained the machine. Then we gave it an input first in the form of a downloaded image and then with the laptop's webcam and observed the results!
At first, the recognition failed. As we see, the input was ginger but the output was bell pepper. That was because we had not put enough training data for ginger so it failed to recognize it.
After inserting more data to the classifier, we tested it again with a picture of zucchini and the results were accurate. There was the limitation that we didn't have any real vegetables in class but still, it managed to capture it.
For the second part of the workshop, we explored this social media emulator called Somekone.
The interface was very similar to Instagram, with images showing up on the feed. We were left to explore the app freely. It had limited interactivity compared to a real social media app but you could do several actions, such as:
Like a post
Comment on a post
Share a post
After some time browsing the app, we collectively observed that the algorithm kept bringing more images of our preferred thematic area based on our likes. For example, if I liked a picture of a cat and then scrolled down or updated my feed, I would get more pictures of cats.
After we were finished with exploring, we saw what the app had recorded about our user behavior.
On the left this image shows the network of connections that the application was building throughout our browsing, based on our preferences when we were interacting with the app. People who liked or commented on the same pictures, people who liked similar thematics were in the end linked to eachother in a node graph.
The purpose of this was to show us in a small scale how social media mechanisms work, how they collect our data and use it for advertising and marketing purposes and so on.
Write a letter to the decision-makers in which you reflect on the operation of social media from your own point of view.
Tell us about your own use of social media (e.g. what you do on social media and why)
Reflect on the following issues based on the lessons that dealt with social media:
advantages of data collection, profiling and recommenders
possible harms, risks and effects of data collection, profiling and recommenders
Present a reasoned solution proposal on how the operating mechanisms of social media could better take into account children and young people as users.
You can also write about other ideas related to social media that you would like adults to take into account.
Dear [Chief Product Officer of Instagram/TikTok name],
I am writing to you as a Master's student of Learning, Education and Technology, highly interested in how technologies and digital tools are affecting people's learning experiences. I am writing regarding a highly relevant topic to your work and my interests, which are the operations of social media and their correlation to young people as users.
Firstly, laying down my own experience with social media, I can say that it has been present in one way or another throughout my early teenage life onwards. Firstly, through Tumblr which was a means of expression and communication with friends, then with Facebook and Instagram, and most recently, TikTok. It gradually took a deeper role in my life, often affecting my emotions based on content that I watched, affecting my views, shaping my self-image. In my adult life, social media has been accompanying me daily, most of the times for entertainment purposes but is also my main source of information about the world. That being said, I think it holds a great weight for me as for the majority of people, therefore the mechanisms behind it, such as data collection, profiling and recommender algorithms, should be discussed further.
On the one hand, personalizing a user's experience by providing recommendations according to their interests, can be beneficial for the user's interaction with the social media application, because it might help them discover similar products, activities and can somehow save time from people trying to find this information themselves. It is an efficient workaround that deals with the problem of information abundance and targets the meaningful content for the users.
On the other hand, there are matters or privacy, consent, and overall freedom of choice regarding users' data being collected. In most cases, there isn't enough transparency by the companies who perform such actions, about the methods they are using, the endpoint to which the data reach, but also the extent of the data collection, or, if there is any relevant information, it is not in an accessible format. This overall creates a justified suspicion, when, simultaneously news are being heard about "data being sold for X amount of money", about "data breaches" in huge companies like Zoom a few years ago, and the list goes on. Moreover, the demographics also is a core point of interest. Specifically, as you already know, in major social media platforms such as Instagram, TikTok that you are working on, the majority of users are young teenagers and children, with the age average dropping lower as the years go by. This changes the whole landscape of how information should be processed and presented to the users.
When referring to children and young people as users, undeniably there is a shared responsibility among parents and the social media platform. The platform itself cannot be held responsible for the uncontrollable exposure on children on it, since it is something that originates from the child's way of living, boundaries that have been set inside the family and active effort from the parents to engage the child in other activities and cultivate interests. However, the people working in lead positions behind those platforms can be, and should be, held accountable for potentially harmful content and for their intentions not being as human-centered as possible. It is no secret that protecting people's feelings or identities is not the priority.
Therefore, as a person studying people's learning, I highly recommend that social media platforms would take into account principles of educational psychology and the multifaceted aspects of one's learning. There are two directions in which my recommendation is based on. The first one is that young people, with their systematic use of social media, unknowingly start to create digital footprints, build on their digital identities and are becoming active contributors to a global digital community. They are experiencing all of this, though, without having any digital literacy skills to base their actions on. So my proposal would be that the instrument of expression/communication/entertainment, which is social media, should also become the instrument for transmitting important information. I believe that data collection results have given plentiful information already about what is it that captures young peoples' attention, so it could be used for building appealing educational material. What remains to be done is for teams like yours to actually utilize this information and collaborate with other experts, building a multi-disciplinary group that studies young people's interactions with social media in different scopes, but with a common goal, to raise their awareness on digital literacy, digital security and safety.
The second direction of my thinking is that children's, especially young teenagers' emotional world is very unstable and fragile. This significant information cannot be omitted. A young person that consumes content on social media is a novice user that is highly influenced by current trends and often follows them mindlessly. Also, children and young people are curious by nature, they might often search for content that displays acts of violence, or other potentially harmful content to their mentality.They are now at the stage of figuring out the world around them, themselves,their self-image, their sexuality, interests, their real identity. Profiling them and presenting recommendations based on their choices alone, without taking into account that the content that they watch and the content that is good for them are two different things, creates unsafe, if not harmful, environments for young people. Additional filtering of the content is essential. Again, this requires collaboration and formation of expert groups that deal with the developmental stages of people based on educational psychology concepts and apply those in social media algorithms.
Closing this letter off, I strongly believe that all people, but especially young ones should not be considered as pawns for companies to making profit out of, but their identities should be protected as valuable members of our societies that are still developing their mental, emotional worlds. The basis of this is reforming the social media companies' intentions and priorities. Institutional knowledge exists, the learning scientists and educational experts are plenty, but the links are missing. I hope that you will take the time to read and consider my proposals.
Best Regards,
Athina Kardiakou
N. Pope, H. Vartiainen, J. Kahila, J. Laru and M. Tedre, "A No-Code AI Education Tool for Learning AI in K-12 by Making Machine Learning-Driven Apps," 2024 IEEE International Conference on Advanced Learning Technologies (ICALT), Nicosia, North Cyprus, Cyprus, 2024, pp. 105-109, doi: 10.1109/ICALT61570.2024.00037.