What is Wrong With This Company?
Facebook is a common target for both government regulation and user complaints. Being around for over a decade, Facebook certainly had its zenith long ago. However, as the world’s legislative bodies slowly catch up to the evolution of the digital realm and begin to recognise users’ rights around the globe, instead of adapting its business model, Facebook seems to choose the second-best option. Instead of adapting their business model, the company apparently attempts to force the globe back into the “good ol’ days” when user data was still just prey lying around to be eaten by a lion from silicon valley.
It took several weeks for me to write this article. Not so much because I don't think that today's topic is important, but rather because I felt I would do injustice to the cited papers here. I still have that vague feeling, but I am convinced that I did my best to honour the work of my fellow researchers. This article deals with methodological scepticism, or, for short: Never trust your methods.
What is a Generator?
This week, we implemented a classifier in our first lab-session of the Natural Language Processing course that I currently take at LiU. To pass the course, we actually have to write a lot of Python code, and as we – the PhD students from the Institute for Analytical Sociology – do not have any formal education in programming and computer science, it is proving hard to receive good results. One question my colleague had during the week was: “What is a generator?” Here’s the answer I gave her. (Probably a better one, because I had a few days to think about it.)
Setting up Python, numpy, and PyTorch natively on Apple M1
The new Apple M1 devices have received quite the attention in the past months. However, data scientists and engineers have been wary of upgrading too soon, and in my opinion rightfully so. However, it is possible to run a development setup natively on the ARM-architecture. In this post, I describe how.
Class Struggle from Above
The main goal for governments globally is to completely drive down daily new cases of COVID. However, the current political strategies prove unable to reduce incidences. While public discourse asks how that could be and lockdowns are being extended, one initiative sheds light onto one sector that has been left almost alone so far.
On Wednesday, violent Trump-supporters, fascists, and white supremacists forced their way into the U.S. capitol building, forcing the senators to interrupt their ongoing meeting and forcing them to evacuate. We could see guards pointing guns at civilians, ransacked offices, and random dudes sitting in the chairs of elected officials. This atrocity highlights three important threads in global political discourse which I comment upon in this post: The demise of the nation state, the fragility of positive law, and, most importantly, how modern fascism is so successful.
Happy New Year 2021
Let's be honest: We all wanted 2020 to end. While I want to stay quiet on the question of whether or not 2021 will actually be a better year, I want to say a little bit about my plans for the year. Right now, a monotonous phase of extracting a lot of information from research literature has started, but the first results are interesting. It seems as if statistics share some commonality with engineering.
Causation, Prediction, Explanation
Sociology is in upheaval: After machine learning technologies brought about huge changes in the area of engineering, social sciences are catching up, introducing more and more machine learning methods into their research. But is machine learning really the next big thing, or rather something additional that you may or may not use? In this blog post I focus on initial research questions that popped up after the first weeks of my PhD, and which will guide my work in the future.
After six weeks in Norrköping, and two weeks before Christmas, it is time for a first feedback. How is it to start a PhD in the mid of a global crisis?
Data Analysis with Python, VS Code and Jupyter
A big part of my PhD will consist of performing data analysis, enhanced with machine learning techniques where appropriate. But before turning to methods such as Latent Dirichlet Allocation (LDA), Word Embeddings, or even model-based classifiers ("neural networks"), it is time to set up my data analysis toolchain again. I have been using Jupyter with VS Code for a few years now, and in this article I explain how you can do the same.