5/5 (2) In the recent years, machine learning technologies are increasing in popularity among variety of different fields. The vast amounts of data allow data scientists to play around and try to answer questions that are impossible to accurately calculate without such technologies. Some of these questions can be highly important and sensitive – for example, deciding whether a loan application will default, or whether an applicant should be considered for the job. There are variety of models that can predict such outcomes, however, is their prediction generalizable and free of biases?

A classical example of this is a neural network classification model, that was built to identify whether in a given picture there is a husky or a wolf. Surprisingly, the model was performing with a high accuracy and predicting the correct animals. However, researchers quickly realized, that something was wrong with it. The model, was basing its decisions purely on one aspect – whether the picture contains snow. Well, that is quite a logical explanation, as pictures of wolves do indeed have usually snow in them, however, given the explanation for this reasoning, I doubt that anyone would want to apply this biased model in a real life situation.

In making predictions, there is often a trade-off between interpretability and model accuracy. When a linear model predicts an outcome, the prediction is just input variables multiplied by different weights – these can be easily explained. However, when using a more advanced model, such as gradient boosting classifier or neural networks, this interpretation becomes complicated.
In recent years, however, there has been some effort to tackle this; one of them is the so called Local Interpretable Model-agnostic Explanations (LIME). Without going too deep in the theoretical details (I will provide some links for further information) LIME is essentially a technique, that helps better explain the reasoning behing any model‘s particular prediction, locally – meaning that it only allows to analyze a small portion of the network/model.

In practice, LIME allows to understand predicitons, by visualizing the features of a given observation that had the most impact in making that particular prediction. Returning back to our wolf/husky identifier – in a given picture, LIME will highlight the snow that that there either is, or is not, hence explaining the behaviour of that particular model.

The importance of this, is that it allows data scientists to better evaluate their models and their applicability. When a model will predict a loan applicant as ‚likely to default‘, and a bank makes the decision to deny him that loan – managers will be able to explain to the person what are the exact reasons. In a critical situations, we cannot simply use black boxes that give us answers to our simplified questions – we also need to know the reasoning.

Sources/Further information:
Video: Interpretable Machine Learning Using LIME Framework – Kasia Kulma (PhD), Data Scientist, Aviva
Article: Guide to Interpretable Machine Learning

5/5 (1) When I first heard about Barbie’s with Artificial Intelligence (AI), I was not sure whether to be celebrating a cool new application of AI or to be concerned. Several questions arose in my head: What kind of things will barbie say or not say? Who is allowed to program the AI? Is there even legislation in place to guide the development of AI in toys for children?

Many institutions and organizations focus on these questions on the ethical considerations in designing technology. For instance, The Center for Humane Technology originates from design ethicist at the biggest tech firms in Silicon Valley, wanting to make a change for the better. They raise concerning points on how technology has shaped our society in negative ways; companies like Facebook successfully manage to fight for a consumers screen time by creating addictive features like scrolling through content, exposure to social media sometimes creates unrealistic expectations and harms self-image, and some primary school children measure their friendship in Snapchat streaks. 

Another example of an institution that focuses on technology’s role in society here in the Netherlands is Waag Society. They organize Hack-a-thons to allow people from multiple backgrounds to come together and think about matters such as the application of AI in society, and how we should shape these in the most beneficial way. You can find interesting perspectives and discussions on their website, such as articles in which they propose to create a few key ‘design principles’ for these emerging applications of technology for children’s toys, to safeguard ethical principles in design and programming whilst legislation may still lag behind.

A third example I would like to share with you is the Next Nature Network, a network of people that uses science, art and education to discuss how we can find new ways to live together with technology. They have interesting publications and events to explore “how to dream, build and live in the next nature”.

Personally, I think that it is good to keep raising questions when you’re involved with new applications of technology, especially now the BIM program gives us the knowledge, tools and frameworks to be shapers and facilitators of these technological transformations.

And what about Barbie? A committee designing the toy has decided on her personality, what she will value and how she will respond. I wonder what her personality is like. If you’re curious, you can see the “Hello Barbie” talk here.

What do you think of a Barbie with AI? Would you buy it for your kids in the future? At the moment, I don’t think I will. But then again, that’s what most people said about the mobile phone in 1999 too.

No ratings yet. Up until now Google Scholar most popular search tool for peer reviewed articles. This might change in the coming years as a system called Semantic Scholar gains traction. Developed by the researchers of the Allen Institute of Artificial Intelligence under the lead of Marie Hagman the system just got another update and now includes more than 40 million papers.

It differs from systems such as Google Scholar because it is not based on citations in other papers or the rate or recurring phrases to determine the relevance of papers it uses AI and bases its results on the context of the phrases which leads to a higher accuracy.

As with most AI systems Semantic Scholar is based on a neural network that the team around Ms. Hagman trained to recognise scientific phrases and their contexts. This was done by giving it manually highlighted topics out of abstracts to develop a pattern at first and then in the second phase give it another batch of abstracts the system then recognises by itself and is afterwards corrected by Ms. Hagmans team.

So, when you start the research for your master thesis give it a try! At least it is nice to throw that buzzword out there: “I did my thesis using AI.”

However, apart from getting the right papers to base your thesis on there are a couple of other topics for which there are some useful and mainly free tools out there. Here is an overview of the most popular ones:

Citavi

Start putting your references and abstracts of papers and books into Citavi, a free reference management software and let it take care of formatting your citations and references. Plus, with the Microsoft Word Citavi plug-in it will lift your writing speed to a new level.

LaTeX

LaTeX is an open-source document preparation system that was designed for scholarly and technical writing, and is great for handling large documents. It is a powerful and highly customizable typesetting system that, in contrast to MS Word, separates the content and document design. However, it is not as straight-forward as Word but if you are lucky EUR might provide pre-set templates that match the Thesis formatting guidelines and the last couple of days you usually spend on formatting you can now spend on planning your holiday after successfully handing it in! So, if interested, maybe check out this great, detailed article on why you should use LaTeX for writing your dissertation.

Pomodoro technique

I first heard of the Pomodoro technique in the MOOC (Massive open online course) by Barbara Oakley on “Learning how to learn“. Which I recommend very much as it provides a lot insight into how to fight procrastination. The simplest and yet most effective technique is the Pomodoro. You just set a timer for a 25-minute session of work. That’s it. The timer goes off after the time and you get a short break (five minutes). After four pomodoros, you get a long break (10 minutes). You’ll be amazed at just how much you can get done in 25 minutes, and how amazing your focus is when you know you’re on a deadline, even if it’s self-imposed. So, take your phone or any other clock set a timer and there you go! And if you earned a little bit of spare time with this technique why not use it to do the MOOC or read the book (A mind for numbers) on which it is based.

Sources:

https://www.economist.com/news/science-and-technology/21730489-get-neural-network-do-it-you-better-way-search-through-scientific

https://www.citavi.com/en/index.html

https://www.coursera.org/learn/learning-how-to-learn

https://thesiswhisperer.com/2011/10/13/how-i-use-technology-in-my-phd/

https://www.insidehighered.com/blogs/gradhacker/best-software-writing-your-dissertation

Why you should LaTeX your dissertation; or, why you don’t have to write your dissertation in Word.