The science of learning updates itself regularly. Accordingly, having already written about “learning how to learn”, here’s a post that features rolling updates on anything broadly related to learning that has caught my attention since then.

The updates appear in ascending order below. However, before we get into that, it’s worth revisiting the value of improving our ability to learn. In this regard, we may take the advice of Tony Yeoh, the Chief Information Officer of Intercontinental Hotels Group who oversees his corporate offices’ technology services and infrastructure.

Unlike some doom-mongers, he does not see the advance of AI, nanotech, and genomic medicine necessarily disrupting the economy and human welfare in dire fashion simply because they are controlled by technocrats whose aptitude and interests may not align with human well-being.

Many of these technologies are here to stay, and they needn’t necessarily have pernicious effects. Yeoh notes that we can optimize our ability to work alongside them (i.e., be good at things intelligent machines aren’t good at), by emphasizing skills like creative expert thinking, complex communication, and pattern recognition across broad, and sometimes incommensurate, domains of knowledge.

To achieve this, he suggests that policymakers consider how education systems can encourage interdisciplinary life-long learning. At the personal level, then, this translates to learning how to learn constantly and effectively, and, by extension, supporting efforts to improve them. 

[10 March 2018]

Since completing the Learning how to Learn MOOC last year, I’ve been a regular recipient of their weekly email updates. If you’ve not done the course but would like a weekly list of resources about learning, these updates are also available on Barbara Oakley’s website. Her list of recommended books is worth looking into as well. The most recent update sent on 8 March links to studies purporting to show how some widespread beliefs about the science of learning are mistaken. I will address them in this post.

To preface these misunderstandings, Oakley recommends David A. Sousa’s How the Brain Learns 5^th edition. Having done so, she offers some caveats about the book. She argues that the latest edition has not engaged with recent and significant criticisms on topics within the science of learning. These include learning styles, stereotype threat, multiple intelligences, and concept mapping. She gives links to some of those studies. Let’s explore each of them.

1. Learning styles

If you’ve ever been in school or read around the topic of learning, you would’ve come across the intuitive idea that learnings styles account for differential learning ability. If you’re a visual learner, you’d learn faster with visual aids than someone who’s an aural or kinesthetic learner. Likewise, if you are a logical learner, you’d pick up propositional logic and mathematics faster than someone who isn’t.

Learning styles loom large in the popular imagination. Their models of learning are recommended by numerous guides, books, and courses, from kindergarten to graduate school. The evidence for them, however, is weak. Many papers can be cited to show how popular understandings of learning styles are predicated on sloppy research and misreadings of the science.

Sloppy research and misreadings happen partly because education researchers don’t have the expertise to interpret neuroscience papers accurately. Learning styles are based on the well-established fact that different regions of the brain (the cortex in particular) engage in visual, auditory, and sensory processing. It’s easy enough to extrapolate from this fact that students learn at different speeds according to their most developed regions. However, as the article observes, interconnectivity within the brain dissolves this idea.

Sloppy research and misreadings also happen because of wishful thinking. Educators who unwittingly oversimplify the neuro-scientific ideas they’ve read are typically driven by a sincere desire to improve educational outcomes. But this pervasive myth, if it indeed remains vindicated as a myth, will only be a disservice to educators and students. It doesn’t help that learning styles are low cost and easy to implement in classrooms.

For more, here are some of the research papers on neuromyths and other myths in higher education.

2. Stereotype threat

A social-scientific study about stereotype threat published in 1995 by Steele & Aronson had a huge impact in the field of psychology: it was cited more than 5000 times. The study purported to show that the long-standing differences in standardized test scores between Caucasians and African Americans could be dissolved by simply framing the test as a challenge rather than a test of ability.

The study assumes that African American students could be performing badly in these tests because they’ve internalized negative stereotypes. The researchers hypothesize that by switching the testing frame from raw ability to a willingness to be challenged, these stereotypes could be momentarily suppressed. This, in turn, should improve testing outcomes. According to their results, this is just what they found.

Many have since touted the power of stereotype threat and the need to remove such stereotypes in order to promote academic equality among the different races in America. The immediate draw of this study is obvious. It bears a heart-warming message about the social causes of academic inequality and it offers a simple solution: frame tests as challenges.     

Problems with the study, however, have been pointed out by social psychologists like Lee Jussim. In his article, Jussim argues that the results from the study are “overstated, overpromoted, and oversold” for the main reason that the SAT scores used as evidence were adjusted with prior SAT scores using an Analysis of Covariance (ANCOVA), a statistical technique.

You may read the link for detailed information about how ANCOVA can unwittingly distort statistical differences. The bottom line is that, without applying the ANCOVA, there is no good evidence for stereotype threat. The pre-existing difference in scoring ability remains even when the frame of the test was shifted to remove stereotype threat.      

3. Multiple Intelligences

Howard Gardner forwarded the theory of multiple intelligences against the prevailing notion that people possessed a kind of general intelligence which powered their abilities, and which could be measured by IQ tests. Beyond verbal and logical intelligence, which he believed IQ could only measure, Gardner argued that people can also have musical, kinesthetic, spatial, interpersonal, intrapersonal, and naturalistic intelligence.

Despite its egalitarian appeal and popularity, there is no good evidence for this theory. Unlike verbal and logical aptitude, which can be measured by IQ tests and which predicts student performance in some subjects, the definitions surrounding Gardner’s types are woolly. This makes them hard to measure.

They can, however, be measured if the requisite definitions are put in place. When researchers did so, they found that people who were good at one kind of intelligence tended to be good at the others. When correlated with the Big Five, researchers found that five of the intelligences were positively correlated with openness to experience. This matches the already known correlation between openness to experience and general intelligence. So, the simple inference from the above results is that there are no independent multiple intelligences, but rather, a general intelligence that applies across ability domains.

People are not just made up of their intelligence. Clearly, traits such as social skills, wisdom, and motivation also matter. However, they should not be considered distinct forms of intelligence since there is no evidence for them and good evidence against them. Teaching according to multiple intelligence is therefore not advised. Educators must rely on evidenced-based teaching methods if they want the best chance of success.

4. Concept Mapping

It is commonly assumed that learning is best achieved with some combination of absorbing knowledge and reflecting on experience. Concept maps, which may be classified among these activities, help the student see the big picture of whatever they are trying to learn. Studies have shown, however, that they may not be as effective for recall and understanding as another learning method: good old retrieval practice.   

Retrieval practice involves the process of reconstructing knowledge, as in a memory test with flash cards for a foreign language timed at crucial points along the subject’s forgetting curve (and others like that). Concerning its relative effectiveness, researchers hypothesize that the effort required to reconstruct knowledge, to organize and home in on specific areas of learned material strengthens understanding and recall. This, in turn, supercharges learning.

Summing up

The educational research space is filled with ideas about effective learning. Some have good evidence to support them, some don’t. The four ideas mentioned are each predicated on some modicum of fact: different regions of the brain govern different tasks; test scores can be affected by negative stereotypes; factors that cannot be measured by IQ can nevertheless contribute to a person’s learning outcomes; and concept maps are a good way to get an overview of a subject.

It does not follow, however, that their effects are anything close to what their popularisers say or believe. Better to just focus on methods that have been proven to be effective, like those Oakley recommends in her course.

[16 Dec 2017]

The YouTube channel Med School Insiders, produced by a physician, details productivity hacks and habits that led to his success in med school. There is some overlap with the Learning How to Learn course.

[8 Oct 2017]

The educational YouTube channel Crash Course started a series some weeks back on Study Skills. It’s hosted by Thomas Frank who has his own channel on the same broad topic. Definitely worth checking out if you haven’t already.

[26 Sept 2017]

In 1997, the IBM supercomputer Deep Blue defeated then world chess champion, Garry Kasparov. Some years before that, Chinook, a computer program, tied with then checkers world champion Marion Tinsley six times before Tinsley withdrew for health reasons.

Since then, artificial intelligence has been making the news by beating humans at their most cherished games. For example, last year, AlphaGo made headlines when it defeated Lee Sedol at Go, a game that’s understood to be several times more complex (in a subjective way) than chess.   

This year, Elon Musk’s OpenAI has developed a self-learning bot for DOTA 2, a video game considered to be several orders of magnitude more subjectively complex than traditional board games (common knowledge for anyone who has played DOTA). And it has so far defeated eight top players. Check out Mashable’s cover of the incredible story.

Will self-learning artificial intelligence bring humanity closer to its extinction, as Terminator creatively predicts, and as some technocrats (like Musk) worry a great deal about; or will it enable humanity to scale new heights, when, for example, we learn from them, like those DOTA players, and learn to work alongside them? Time will tell. Though it might be best to find out what automation will likely mean for our livelihoods in the meantime. 

[24 Sept 2017]

Michael Boyd, a Scottish YouTuber, posts popular video dairies of himself learning new skills. They range from solving a Rubik’s Cube in under two minutes with sixteen hours of practice to stacking ten dice after five-plus hours of practice. He has been posting these videos every month for the past two years. His videos are inspiring. They show the amazing things that a can-do attitude, persistence, and a growth mindset can achieve. Here’s his channel.

[23 Sept 2017] 

Ulrich Boser, senior fellow at American Progress, published a book this year titled Learn Better. Like the Learning How to Learn course, Boser explains how the science of learning can help people gain expertise quickly. Here are some of his insights.  

We can get better at learning with specific tools and strategies. The more confident you are in something, when you are proven wrong, you are more likely to have learned and retained that information for the future.

Learning is therefore sense-making, and the analogy with a computer storing data is not a good one. This means that many old-fashioned learning strategies like re-reading, highlighting, and concept-mapping, are not as effective as self-quizzing, self-explaining, and explaining to someone else. In other words, recalling the concept is the most effective way to learning it.