Article | May 3, 2021
Clear conceptualization, taxonomies, categories, criteria, properties when solving complex real-life contextualized problems is non-negotiable, a “must” to unveil the hidden potential of NPL impacting on the transparency of a model.
It is common knowledge that many authors and researchers in the field of natural language processing (NLP) and machine learning (ML) are prone to use explainability and interpretability interchangeably, which from the start constitutes a fallacy. They do not mean the same, even when looking for a definition from different perspectives.
A formal definition of what explanation, explainable, explainability mean can be traced to social science, psychology, hermeneutics, philosophy, physics and biology. In The Nature of Explanation, Craik (1967:7) states that “explanations are not purely subjective things; they win general approval or have to be withdrawn in the face of evidence or criticism.” Moreover, the power of explanation means the power of insight and anticipation and why one explanation is satisfactory involves a prior question why any explanation at all should be satisfactory or in machine learning terminology how a model is performant in different contextual situations. Besides its utilitarian value, that impulse to resolve a problem whether or not (in the end) there is a practical application and which will be verified or disapproved in the course of time, explanations should be “meaningful”.
We come across explanations every day. Perhaps the most common are reason-giving ones. Before advancing in the realm of ExNLP, it is crucial to conceptualize what constitutes an explanation. Miller (2017) considered explanations as “social interactions between the explainer and explainee”, therefore the social context has a significant impact in the actual content of an explanation. Explanations in general terms, seek to answer the why type of question. There is a need for justification. According to Bengtsson (2003) “we will accept an explanation when we feel satisfied that the explanans reaches what we already hold to be true of the explanandum”, (being the explanandum a statement that describes the phenomenon to be explained (it is a description, not the phenomenon itself) and the explanan at least two sets of statements, used for the purpose of elucidating the phenomenon).
In discourse theory (my approach), it is important to highlight that there is a correlation between understanding and explanation, first and foremost. Both are articulated although they belong to different paradigmatic fields. This dichotomous pair is perceived as a duality, which represents an irreducible form of intelligibility.
When there are observable external facts subject to empirical validation, systematicity, subordination to hypothetic procedures then we can say that we explain. An explanation is inscribed in the analytical domain, the realm of rules, laws and structures. When we explain we display propositions and meaning. But we do not explain in a vacuum. The contextual situation permeates the content of an explanation, in other words, explanation is an epistemic activity: it can only relate things described or conceptualized in a certain way. Explanations are answers to questions in the form: why fact, which most authors agree upon.
Understanding can mean a number of things in different contexts. According to Ricoeur “understanding precedes, accompanies and swathes an explanation, and an explanation analytically develops understanding.” Following this line of thought, when we understand we grasp or perceive the chain of partial senses as a whole in a single act of synthesis. Originally, belonging to the field of the so-called human science, then, understanding refers to a circular process and it is directed to the intentional unit of discourse whereas an explanation is oriented to the analytical structure of a discourse.
Now, to ground any discussion on what interpretation is, it is crucial to highlight that the concept of interpretation opposes the concept of explanation. They cannot be used interchangeably. If considered as a unit, they composed what is called une combinaison éprouvé (a contrasted dichotomy). Besides, in dissecting both definitions we will see that the agent that performs the explanation differs from the one that produce the interpretation.
At present there is a challenge of defining—and evaluating—what constitutes a quality interpretation. Linguistically speaking, “interpretation” is the complete process that encompasses understanding and explanation. It is true that there is more than one way to interprete an explanation (and then, an explanation of a prediction) but it is also true that there is a limited number of possible explanations if not a unique one since they are contextualized. And it is also true that an interpretation must not only be plausible, but more plausible than another interpretation. Of course there are certain criteria to solve this conflict. And to prove that an interpretation is more plausible based on an explanation or the knowledge could be related to the logic of validation rather than to the logic of subjective probability.
Narrowing it down
How are these concepts transferred from theory to praxis? What is the importance of the "interpretability" of an explainable model? What do we call a "good" explainable model? What constitutes a "good explanation"? These are some of the many questions that researchers from both academia and industry are still trying to answer.
In the realm on machine learning current approaches conceptualize interpretation in a rather ad-hoc manner, motivated by practical use cases and applications. Some suggest model interpretability as a remedy, but only a few are able to articulate precisely what interpretability means or why it is important. Hence more, most in the research community and industry use this term as synonym of explainability, which is certainly not. They are not overlapping terms. Needless to say, in most cases technical descriptions of interpretable models are diverse and occasionally discordant.
A model is better interpretable than another model if its decisions are easier for a human to comprehend than decisions from the other model (Molnar, 2021). For a model to be interpretable (being interpretable the quality of the model), the information conferred by an interpretation may be useful. Thus, one purpose of interpretations may be to convey useful information of any kind. In Molnar’s words the higher the interpretability of a machine learning model, the easier it is for someone to comprehend why certain decisions or predictions have been made.” I will make an observation here and add “the higher the interpretability of an explainable machine learning model”. Luo et. al. (2021) defines “interpretability as ‘the ability [of a model] to explain or to present [its predictions] in understandable terms to a human.” Notice that in this definition the author includes “understanding” as part of the definition, giving the idea of completeness. Thus, the triadic closure explanation-understanding-interpretation is fulfilled, in which the explainer and interpretant (the agents) belong to different instances and where interpretation allows the extraction and formation of additional knowledge captured by the explainable model.
Now are the models inherently interpretable? Well, it is more a matter of selecting the methods of achieving interpretability: by (a) interpreting existing models via post-hoc techniques, or (b) designing inherently interpretable models, which claim to provide more faithful interpretations than post-hoc interpretation of blackbox models. The difference also lies in the agency –like I said before– , and how in one case interpretation may affect the explanation process, that is model’s inner working or just include natural language explanations of learned representations or models.
Article | May 3, 2021
Does the success of companies like Google depend on that of the algorithms or that of data? Today’s fascination with artificial intelligence (AI) reflects both our appetite for data and our excitement about the new opportunities in machine learning. Amalio Telenti, Chief Data Scientist and Head of Computational Biology at Vir Biotechnology Inc. argue that newcomers to the field of data science are blinded by the shiny object of magical algorithms and that they forget the critical infrastructures that are needed to create and to manage data in the first place.Data management and infrastructures are the little ugly duckling of data science but they are necessary for a successful program and therefore need to be built with purpose. This requires careful consideration of strategies for data capture, storage of raw and processed data and instruments for retrieval. Beyond the virtues of analysis, there are also the benefits of facilitated retrieval. While there are many solutions for visualization of corporate or industrial data, there is still a need for flexible retrieval tools in the form of search engines that query the diverse sources and forms of data and information that are generated at a given company or institution.
Article | May 3, 2021
Machine Learning (ML) has taken strides over the past few years, establishing its place in data analytics. In particular, ML has become a cornerstone in data science, alongside data wrangling, and data visualization, among other facets of the field. Yet, we observe many organizations still hesitant when allocating a budget for it in their data pipelines. The data engineer role seems to attract lots of attention, but few companies leverage the machine learning expert/engineer. Could it be that ML can add value to other enterprises too? Let's find out by clarifying certain concepts.
What Machine Learning is
So that we are all on the same page, let's look at a down-to-earth definition of ML that you can include in a company meeting, a report, or even within an email to a colleague who isn't in this field. Investopedia defines ML as "the concept that a computer program can learn and adapt to new data without human intervention." In other words, if your machine (be it a computer, a smartphone, or even a smart device) can learn on its own, using some specialized software, then it's under the ML umbrella. It's important to note that ML is also a stand-alone field of research, predating most AI systems, even if the two are linked, as we'll see later on.
How Machine Learning is different from Statistics
It's also important to note that ML is different from Statistics, even if some people like to view the former as an extension of the latter. However, there is a fundamental difference that most people aren't aware of yet. Namely, ML is data-driven while Statistics is, for the most part, model-driven. This statement means that most Stats-based inferences are made by assuming a particular distribution in the data, or the interactions of different variables, and making predictions based on our mathematical models of these distributions. ML may employ distributions in some niche cases, but for the most part, it looks at data as-is, without making any assumptions about it.
Machine Learning’s role in data science work
Let’s now get to the crux of the matter and explore how ML can be a significant value-add to a data science pipeline. First of all, ML can potentially offer better predictions than most Stats models in terms of accuracy, F1 score, etc. Also, ML can work alongside existing models to form model ensembles that can tackle the problems more effectively. Additionally, if transparency is important to the project stakeholders, there are ML-based options for offering some insight as to what variables are important in the data at hand, for making predictions based on it. Moreover, ML is more parametrized, meaning that you can tweak an ML model more, adapting it to the data you have and ensuring more robustness (i.e., reliability). Finally, you can learn ML without needing a Math degree or any other formal training. The latter, however, may prove useful, if you wish to delve deeper into the topic and develop your own models. This innovation potential is a significant aspect of ML since it's not as easy to develop new models in Stats (unless you are an experienced Statistics researcher) or even in AI. Besides, there are a bunch of various "heuristics" that are part of the ML group of algorithms, facilitating your data science work, regardless of what predictive model you end up using.
Machine Learning and AI
Many people conflate ML with AI these days. This confusion is partly because many ML models involve artificial neural networks (ANNs) which are the most modern manifestation of AI. Also, many AI systems are employed in ML tasks, so they are referred to as ML systems since AI can be a bit generic as a term. However, not all ML algorithms are AI-related, nor are all AI algorithms under the ML umbrella. This distinction is of import because certain limitations of AI systems (e.g., the need for lots and lots of data) don't apply to most ML models, while AI systems tend to be more time-consuming and resource-heavy than the average ML one. There are several ML algorithms you can use without breaking the bank and derive value from your data through them. Then, if you find that you need something better, in terms of accuracy, you can explore AI-based ones. Keep in mind, however, that some ML models (e.g., Decision Trees, Random Forests, etc.) offer some transparency, while the vast majority of AI ones are black boxes.
Learning more about the topic
Naturally, it's hard to do this topic justice in a single article. It is so vast that someone can write a book on it! That's what I've done earlier this year, through the Technics Publications publishing house. You can learn more about this topic via this book, which is titled Julia for Machine Learning(Julia is a modern programming language used in data science, among other fields, and it's popular among various technical professionals). Feel free to check it out and explore how you can use ML in your work. Cheers!
Article | May 3, 2021
In 2020, the gaming market generated over 177 billion dollars, marking an astounding 23% growth from 2019. While it may be incredible how much revenue the industry develops, what’s more impressive is the massive amount of data generated by today’s games.
There are more than 2 billion gamers globally, generating over 50 terabytes of data each day. The largest game companies in the world can host 2.5 billion unique gaming sessions in a single month and host 50 billion minutes of gameplay in the same period.
The gaming industry and big data are intrinsically linked. Companies that develop capabilities in using that data to understand their customers will have a sizable advantage in the future. But doing this comes with its own unique challenges.
Games have many permutations, with different game types, devices, user segments, and monetization models. Traditional analytics approaches, which rely on manual processes and interventions by operators viewing dashboards, are insufficient in the face of the sheer volume of complex data generated by games.
Unchecked issues lead to costly incidents or missed opportunities that can significantly impact the user experience or the company’s bottom line. That’s why many leading gaming companies are turning to AI and Machine Learning to address these challenges.
Gaming Analytics AI
Gaming companies have all the data they need to understand who their users are, how they engage with the product, and whether they are likely to churn. The challenge is gaining valuable business insights into the data and taking action before opportunities pass and users leave the game.
AI/ML helps bridge this gap by providing real-time, actionable insights on near limitless data streams so companies can design around these analytics and act more quickly to resolve issues. There are two fundamental categories that companies should hone in on to make the best use of their gaming data:
The revenue generating opportunities in the gaming industry is one reason it’s a highly competitive market. Keeping gamers engaged requires emphasizing the user experience and continuous delivery of high-quality content personalized to a company’s most valued customers.
Customer Engagement and User Experience
Graphics and creative storylines are still vital, and performance issues, in particular, can be a killer for user enjoyment and drive churn. But with a market this competitive, it might not be enough to focus strictly on these issues.
Games can get an edge on the competition by investing in gaming AI analytics to understand user behaviors, likes, dislikes, seasonality impacts and even hone in on what makes them churn or come back to the game after a break.
AI-powered business monitoring solutions deliver value to the customer experience and create actionable insights to drive future business decisions and game designs to acquire new customers and prevent churn.
AI-Enhanced Monetization and Targeted Advertising
All games need a way to monetize. It’s especially true in today’s market, where users expect games to always be on and regularly deliver new content and features. A complex combination of factors influences how monetization practices and models enhance or detract from a user’s experience with a game.
When monetization frustrates users, it’s typically because of aggressive, irrelevant advertising campaigns or models that aren’t well suited to the game itself or its core players. Observe the most successful products in the market, and one thing you will consistently see is highly targeted interactions.
Developers can use metrics gleaned from AI analytics combined with performance marketing to appeal to their existing users and acquire new customers. With AI/ML, games can use personalized ads that cater to users’ or user segments’ behavior in real-time, optimizing the gaming experience and improving monetization outcomes.
Using AI based solutions, gaming studios can also quickly identify growth opportunities and trends with real-time insight into high performing monetization models and promotions.
Mobile Gaming Company Reduces Revenue Losses from Technical Incident
One mobile gaming company suffered a massive loss when a bug in a software update disrupted a marketing promotion in progress. The promotion involved automatically pushing special offers and opportunities for in-app purchases across various gaming and marketing channels. When a bug in an update disrupted the promotions process, the analytics team couldn’t take immediate action because they were unaware of the issue.
Their monitoring process was ad hoc, relying on the manual review of multiple dashboards, and unfortunately, by the time they discovered the problem, it was too late. The result was a massive loss for the company – a loss of users, a loss of installations, and in the end, more than 15% revenue loss from in-app purchases.
The company needed a more efficient and timely way to track its cross-promotional metrics, installations, and revenue. A machine learning-based approach, like Anodot’s AI-powered gaming analytics, provides notifications in real-time to quickly find and react to any breakdowns in the system and would have prevented the worst of the impacts.
Anodot’s AI-Powered Analytics for Gaming
The difference between success and failure is how companies respond to the ocean of data generated by their games and their users. Anodot’s AI-powered Gaming Analytics solutions can learn expected behavior in the complex gaming universe across all permutations of gaming, including devices, levels, user segments, pricing, and ads.
Anodot’s Gaming AI platform is specifically designed to monitor millions of gaming metrics and help ensure a seamless gaming experience. Anodot monitors every critical metric and establishes a baseline of standard behavior patterns to quickly alert teams to anomalies that might represent issues or opportunities.
Analytics teams see how new features impact user behavior, with clear, contextual alerts for spikes, drops, purchases, and app store reviews without the need to comb over dashboards trying to find helpful information.
The online gaming space represents one of the more recent areas where rapid data collection and analysis can provide a competitive differentiation. Studios using AI powered analytics will keep themselves and their players ahead of the game.