August 2018
Text by Ray Gallon

Image: © metamorworks/

Ray Gallon is co-founder of The Transformation Society. He has over 40 years of experience as a communicator, first as an award-winning radio producer, then in the technical content industries. Ray is the current president of the Information 4.0 Consortium. He is a speaker at events throughout the world, and has contributed to many books and periodicals, most recently as editor of The Language of Technical Communication (XML Press).

Twitter: @RayGallon, @TransformSoc



Learning reinvented: How AI turns technical communicators into educators

The world is changing with increasing speed. So how can we keep up? Lifelong learning has become an absolute necessity, and technical communicators hold a crucial position in enabling an individualized, customized transfer of knowledge.

In March 2018, my collaborator Neus Lorenzo and I had the privilege of hosting a symposium and a workshop at the annual Mobile Learning Week, an event co-sponsored by UNESCO and the International Telecommunication Union (ITU). UNESCO is the educational, scientific, and cultural organization of the United Nations. The ITU, also a UN agency, coordinates telecommunications, spectrum allocations, and policy positions on information and communication services which, it seems, the world no longer knows how to live without.

The event made for an amazing week, during which we were able to interact with some of the smartest people from all over the world on subjects connected to all kinds of learning in so many different cultural contexts, but all associated with the major question of mobility.

A subtheme that ran through many of the interventions, including our own, was Artificial Intelligence (AI). And it’s no wonder – if you look closely at AI, it’s all about learning. A Forbes article by Gill Press (January 2017) listed these as the top ten AI technologies:

  1. Natural language generation
  2. Speech recognition
  3. Virtual agents
  4. Machine learning platforms
  5. AI-optimized hardware
  6. Decision management
  7. Deep learning platforms
  8. Biometrics
  9. Robotic process automation
  10. Text analytics and NLP

With the possible exceptions of hardware and process automation, all of the above involve some type of machine learning or linguistic process. Although machines learn in a different fashion than humans, learning is at the heart of AI. It is also at the heart of technical communication. If you think about it, the different types of user assistance – be it onboarding, task-related, reference material, conceptual information, or anything else – are designed to help users learn about the products they need to use.

One of the most striking takeaways from the Mobile Learning Week was not only how much interest there is in Information and Communication Technologies (ICT) in education, but just how much convergence there is between different areas of education (formal, vocational, informal, non-formal, eLearning, mLearning, training, etc.) and technical communication.

Converging with cognitive science

Ever since the late 80’s, when John Carroll developed his first notions of minimalist information design, technical communicators have been applying principles of cognitive science to the process of learning about technological products. In 2013, I presented a series of webinars on A Cognitive Design for User Assistance for Adobe Technical Communication that drew on the work of different researchers and learning theories to help maximize utility and retention of user guidance.

Concurrently, some researchers were also using cognitive science to develop machine learning. Palm Pilot inventor Jeff Hawkins had been exploring this idea for a long time (see his 2004 book On Intelligence, co-authored with Sandra Blakeslee). He developed a unified theory of the brain that argues that the key to the brain and intelligence is the ability to make predictions about the world by recognizing patterns. He argues that to actually make Artificial Intelligence, all we need to do is teach the machine to find and use patterns, rather than try and teach it to perform specific tasks. This is, in fact, what many machine learning programs do, acquiring massive quantities of training information from Big Data on the Internet. Hawkins calls this a memory prediction system. He claims that such a system is implemented in the brain's cortex and that it is the basis of human intelligence.

Minimalist information design postulates that the quickest way to become productive is to learn by doing. Researcher Roger C. Schank analyzed what happens when we learn in this manner, and he suggests that we acquire “scenes” or tasks that can be generalized from one operational situation to another: We learn to use a credit card to pay for a meal in a restaurant and can then reuse this knowledge to pay the taxi driver who takes us home. Similarly, we learn how to configure parameters when reading a two-dimensional digital medical X-ray, and we can extend that knowledge to configuring new parameters when we want to read a three-dimensional X-ray using software from the same developer.

Schank refers to these situational scripts as Memory Organization Packets (MOPs). Whenever we learn a scene in one MOP, we should be able to generalize it and transfer it to other MOPs in which this scene is applicable. We should also be able to combine scenes that we have learned, in order to create our own new MOPs – and that is at the heart of real learning. Schank is now working on AI, and regularly delivers ideas that run counter to the mainstream thinking in machine learning research.

AI researcher Carlos E. Perez, a keynote speaker at the 2018 Information Energy conference in Amsterdam, is working on the idea of an "intuition machine" that should also be able to transfer what it learns from one context to another.

It should come as no surprise that educators are also calling upon cognitive scientists to better understand how children and adults learn. This is especially important due to the fact that lifelong learning has today become a necessity. As the rate of change in the world accelerates, the labor market is destroying old, familiar jobs while creating new professions that never existed before. Just as an example, the French minister of education, Jean-Michel Blanquer, has named cognitive scientist Stanislas Dehaene to head the National Educational Science Council (Conseil scientifique de l'Education nationale), one of the top educational policy organs in France.

Learning to learn

These days, we hear a lot about how we need to educate children today to take on jobs that don’t yet exist. What we don’t talk about too much is helping adults to make the transition into the world of Industry 4.0, where machines make decisions without human intervention – decisions that affect all of us. This transition for adults is also a critical job for educators.

David H. Autor and Brendan Price of Massachusetts Institute of Technology (MIT) have studied trends in demand for worker skills in the U.S. Their work clearly shows that the only tasks for which demand is rising are those that require high-level cognitive functioning and social (i.e. collaborative) interaction (see Figure 1). It seems logical that results for the rest of the developed world would be similar.

Figure 1: The U.S. labor market shows increasing demand only for advanced cognitive activities
Autor & Price, MIT


In other words, the types of tasks most kids perform in schools today, i.e., routine cognitive tasks, are no longer in demand on the job market. Autor and Price tell us that what will be needed in the future are skills in activities that require problem-solving, intuition, persuasion, and creativity. These include hypothesis testing, diagnosing, analyzing, writing, persuading, and managing people. These are typical skills in educational, managerial, technical, and creative professions such as science, engineering, law, medicine, design, and marketing. These types of activities will be facilitated and complemented by computers, including AI applications, but not replaced by them.

The World Economic Forum projects out to 2020, and confirms the trends observed by Autor and Price (Figure 2).

Figure 2: Change in demand for core work-related skills, 2015-2020, all industries
Source: World Economic Forum Future of Jobs Report


The black boxes to the left of the bars show the percentage of all jobs that will require the given skill set. The top demand is for:

  • Complex problem-solving
  • Social skills
  • Process skills
  • Systems skills
  • Cognitive abilities

Note that content skills are forecasted to be required in only ten percent of all jobs between now and 2020.

Does this mean that there will be less and less work for technical communicators? I don’t think so. But it does mean that our roles will change – indeed, they are already changing. Many of us may be introverts, but we are called upon to collaborate in teams – for Agile development, and for collaborative design processes that can involve design thinking and other process-related methodologies. We design architectures for information creation, curation, and delivery systems, and we work (or will work) collaboratively with customers, product managers, and engineers to design contextualized, personalized information that enables our customers to become experts in the use of our products.

In schools, subject matter learning has become secondary to learning to learn, and to developing critical thinking to make useful evaluations of what students find on the Internet. In technical communication, our paper manuals and static subject matter knowledge are also giving way to notions such as onboarding, information experience, and dynamic delivery, as well as learning by doing.

Artificial intelligence is going to help us deliver information that is highly tuned to the exact circumstances of the user, and even his emotional state at the moment. We already chunk information in structured authoring systems, but we will need to make these chunks even smaller to facilitate rapid updating and translation and to combine and recombine the chunks based on the user’s context.

Such information responds to a user’s immediate, contingent need. The consequence is that the user will often have an isolated piece of very advanced information, which solves an immediate problem, but lacks the background fundamentals that a user would have gotten in a more traditional, linear learning context. We have already become used to this type of information delivery through the use of search engines to answer our contingent needs; we all have "black holes" in our knowledge.

From an information provider’s point of view, the problem is that we don’t want to tell people what they already know, but every user’s black holes are different. The response to this problem should be to stop thinking about delivering static content and provide a well-chosen selection of information that gives the user a wide enough choice to find what is needed, but not so wide as to be paralyzing. The technologies to help us do this are still under development, but they will soon arrive in our toolkits. Taxonomies already exist, and Artificial Intelligence will help us create more useful ontologies in our information offer. Researchers are working on how to describe a user’s context in metadata, so as to create a standard context description metalanguage. Developments that we looked at as "wishful science fiction" only a few years ago are about to become reality.

Are we ready?

Our challenge – both as educators in schools and learning facilitators in the world of technology – is not only to help people learn how to use new tools and new working environments, it’s also to help people learn to live in a society where these technologies are embedded in everyday life. Let’s take one example: A child grows up with Augmented Reality all around him. He plays games that use it; he learns with it in school. He’s had it around since infancy. What notion does that child have about what is real? How can we guide him toward a new understanding of this notion? And how do we guide adults to enter into this realm when they do not have the experience of growing up with it?

Many people have been asking about ethics in the implementation of AI. To what extent will we technical communicators be called upon to write about, and describe, ethical aspects of the technologies we cover? What roles and responsibilities will we have? How will our discipline be impacted by changes in insurance, liability laws, and the risk of increasing litigation as autonomous technologies decide about our lives?

In 1996, the European Commission under Jacques Delors produced a report proposing an integrated vision of education based on the concept that lifelong learning was no longer an option, and four pillars of education:

  • Learning to know – a broad general knowledge with the opportunity to work in depth on a small number of subjects.
  • Learning to do – to acquire not only occupational skills but also the competence to deal with many situations and to work in teams.
  • Learning to be – to develop one's personality and to be able to act with growing autonomy, judgment and personal responsibility.
  • Learning to live together – by developing an understanding of other people and an appreciation of interdependence.

In 2015, the United Nations adopted as official policy a set of 17 Sustainable Development Goals (SDGs) to be attained by 2030. Two of these relate to education and to inclusive and sustainable economic growth. Others relate to energy, infrastructure, climate change, and so on. One of the targets for education is,

By 2030, substantially increase the number of youth and adults who have relevant skills, including technical and vocational skills, for employment, decent jobs and entrepreneurship.

The indicators for attaining that goal are listed as:

Proportion of youth and adults with information and communications technology (ICT) skills, by type of skill.

Couple that with this target for economic growth:

Promote development-oriented policies that support productive activities, decent job creation, entrepreneurship, creativity and innovation, and encourage the formalization and growth of micro-, small- and medium-sized enterprises, including through access to financial services.

It’s easy to see that these goals are interconnected, and learning plays a part everywhere. Taking this into account, and given all that is at stake for human society, we can revise Delors’ four pillars of education to the following:

  • Learning to learn
  • Learning to interact
  • Learning to do with what you learn
  • Learning to be engaged with the common good

These apply whether offering fundamental education to children at school, higher education to university and other post-secondary students, or contingent learning to users of technological equipment that is becoming more and more interactive and autonomous.

One thing is sure: If these technologies cannot help us solve human problems, they will themselves become a world problem we’ll all have to solve.


Additional reading