Big data, machine learning, automated analysis and other chatbots, so many of these concepts have a variety of definitions, but it’s imperative that we clarify what we mean by them before we launch into the exploration of artificial intelligence in the world of recruitment.
Today we being our exploration into this major subject of artificial intelligence. As with any voyage, we must first equip ourselves with the necessary tools to help us on our way, with a map for orientation, and with statistics and studies to make sure we avoid potential hazards. Even so, the paths may still be slippery the roads unpaved, but this is unavoidable when launching yourself into the unknown. In this induction, we invite you to enter into the lexical field of artificial intelligence.
The 11 Indispensable Terms of AI
It’s what it says on the tin.Big data is defined as a collection of data which has become so vast that traditional information management tools are often incapable of processing it.
In HR language: as someone working in human resources, you’re responsible for dealing with large amounts of data shared with your company; personal data of employees, salaries, performance reviews; all this big data is entrusted to you!
AI is a collection of theories and techniques compiled with the view of making machines capable of imitating human intelligence, for example making decisions in spite of an uncertain environment.
In HR language: artificial intelligence is a new HR assistant which could carry out tasks which may or may not require human decision, such as automatically sorting CVs, communicating with candidates and inviting the best candidates to continue with their application…
We often speak of ‘learning by example’, and it’s this concept which forms the basis for machine learning, which represents a collection of processes which allow a machine to learn according to a set of rules defined by algorithms which have been formulated by observing human example. Such methods of learning equip a machine to make effective decisions.
In HR language: When training, we always start by introducing a theory, giving examples, applying rules and sometimes exceptions to the rules. Then we put it into practice, and the more you practice the more you learn. Machine learning works in exactly the same way!
Deriving from machine learning, “deep learning” has become the name for the method which consists of making a machine teach itself to handle a large volume of data in a hierarchical manner. The system begins by analysing the data in a simple fashion and will then make its analysis increasingly complex until it produces extremely well refined results.
In HR language: during an interview, you will be able to detect basic social signals right from the beginning; you recognize these small communicative elements which quickly awaken a certain curiosity within you, such as a stammer, a frown or a badly chosen word. Then you can accumulate these signals, associate them with one another and use them to make a decision regarding recruitment. Deep learning can reproduce such abilities.
A flowchart is a graphical representation of rules which lead you to make a decision following a certain system. Each box represents the different possible decisions, which in turn lead to further decisions, until you finally reach the end conclusion. Each of the branches are linked by probability.
In HR language: in recruitment, as you are sorting CVs, you apply certain criteria filters relevant to the role concerned. For example, for the role of an international commercial engineer, you will look to see if the candidate has an engineering degree. If yes, you’ll continue, but if not, the CV will be rejected. You might next look to see if the candidate speaks English, if yes, the CV will be submitted for pre-selection, but if not it will be impossible for this candidate to integrate into the company. Do they have the necessary experience? And so on, up until the point at which you decide to invite the candidate for interview.
Supervised or Unsupervised Learning
The distinguishing factor between these two methods is whether or not human intervention is required during the course of the machine’s learning process. With supervised learning, a human annotates or classifies the data in order to create samples which will guide the machine. Unsupervised learning means that the computer must learn this itself by analysing a large volume of existing data patterns.
In HR language: as a recruiter, you undoubtedly have hundreds of hiring processes under your belt: from sorting CVs to selecting the successful candidate. Within supervised learning, an algorithm will be formed from all these examples in an attempt to learn the logic of your decision-making process, and then to replicate it. On the other hand, unsupervised learning doesn’t depend on your knowledge as a recruiter and will attempt to classify CVs into different categories by itself (commercial profile, technical profile, junior, senior…). The risk here is that the machine could create categories which are irrelevant from a recruitment point of view.
These are algorithms which allow machines to schematically imitate the networks of biological neurons, initially designed with the aim of modelling information processing. These operational rules are based on statistics and improve gradually as they encounter a greater number of situations. For example, they are applied in the recognition of shapes and images, on stock markets or even in medical diagnoses.
In HR language: in recruitment, upon receiving an application, for example, you process certain pieces of information (qualifications, experience, technical competence…). In this process, your neurons apply mysterious mathematical formulae based on your judgements- whether conscious or sub-conscious, which lead to you making a decision: whether or not to continue the application process.
Natural language processing (NLP)
This is the area of artificial intelligence which aims to model and reproduce human language, including understanding and generating words. This is the crossroads between linguistics and IT.
The goal of this is the description, explanation and simulation of thought mechanisms and the human consciousness. This forms a discipline which binds together psychology, linguistics, artificial intelligence and neuroscience, as well as anthropology and philosophy.
A chatbot is an interface posed as an intelligent agent which manages interactions between humans and machines. A software robot of this kind is capable of participating in dialogue. It can ask questions and give responses by following a set of predefined rules. Chatbots can be based on NLP and/or a flowchart.
In HR language: HR can be assisted by machines which use good interpretation of language to collect information about candidates, ask selection questions and respond to precise questions, which can then be sent directly to the relevant parties.
This concerns the application of statistics and artificial intelligence methods used to predict future events or the evolution of variables. It is based on the predictive hypothesis which assumes that a situation will encounter the same evolution as multiple other similar situations have done in the past.
In HR language: predictive analysis could allow you to anticipate recruitments from one year to another based on seasonality and turnover rate. This information would allow you to select the best possible candidates for your business from an application pool.
How are these Concepts Connected?
Now that you know a little more regarding the meanings of each of these common artificial intelligence terms, here’s how they’re all tied together: AI brings about machine learning, deep learning and decision-making flowcharts. All AI is either supervised or unsupervised, and uses algorithms, in particular neurone networks, which are the basis of deep learning. In the same way, AI applications such as chatbots or predictive analysis depend on data, which leads us to big data. Without big data, deep learning would not be very precise. We’ve come full circle! It’s not without reason that the web giants (such as GAFAM and BATX) who deal with big data are today’s masters of AI.