New tool a 'dopamine engine for writing'

New tool a

Despite writing being a critical skill, there has been a consistent drop in students’ proficiency across various year levels. Moreover, many new teachers are entering classrooms feeling ill-equipped to teach this important skill.

According to The Australian Writing Survey (AWS), conducted by the Institute for Learning Sciences and Teacher Education, nearly half (49%) of teachers feel underprepared to teach writing upon completing their Initial Teacher Education (ITE).

Indeed, these issues are not new, and work has been underway for years to help schools lift students’ writing outcomes.

Nearly a decade ago, the Gates Foundation funded a research project to help students master writing. The project led to the creation of a “voice gym” based on cognitive load theory. While this prototype initially missed the mark with younger students who struggled to bridge knowledge gaps and sustain engagement, it was refined over seven years as new features were tested and feedback was gathered.

The breakthrough came with ‘Frankenstories’ – an “anti-AI” interactive tool that motivates students to write. Now an award-winning “unique hivemind experience” being leveraged by tens of thousands of Australian school students, the initiative has only gone from strength to strength.

Below, The Educator speaks to Andrew Duval, co-founder of Writelike to find out more about Frankenstories, how it stands out from other programs aimed at lifting students’ writing outcomes, and the impact it’s having in Australian classrooms.

TE: What are the main factors that motivated the development of Frankenstories?

Essentially, we wanted to create a dopamine engine for writing. Writing is a critical and complex skill but it demands a lot of attention and effort from students without necessarily giving much back in the way of immediate reward. We wanted to create an engagement engine, and that led to the development of Frankenstories, which is a writing game that combines surprise, pressure, and pleasure to make students want to write and—importantly—want to get better at writing.

TE: Can you tell us about the evidence that Frankenstories is grounded in, and what its presence in Australian schools currently looks like?

Frankenstories is a kind of educational psychology onion, with reinforcing layers. The first layer is straight Skinnerian behaviourism in the form of varied rewards on a variable schedule—this builds student engagement. The next layer is genre-based pedagogy—players use authentic text patterns to influence an audience. The third layer is cognitive load theory: small chunks, massed practice, learning from examples (in this case, other students’ responses). The final layer is social learning theory—authentic participation in a communal task, cultural goals and values, learning from proximal peers.

I like to think that we don’t merely apply an evidence base, but also actively contribute to it. We have collaborative relationships with academics at a range of institutions including UNSW, ANU, USyd, UQ and ASU (in the States), and we continually articulate and test new ways of thinking about writing and pedagogy. (Speaking of which, I’ll be presenting some ideas on cross-genre teaching at VATE at the end of November.)

Currently, we have around 31,000 students playing each month. 70% of those are in the US and Canada, and 30% in Australia and NZ. We have teachers in all states in Australia, in all school contexts: state, Catholic, private, urban, regional, remote. And we have classes playing the same underlying game from grades 4-12.

TE: I understand that Frankenstories recently won a prestigious AI grant from MIT Solve, despite being what you call “an anti-AI” product. Can you explain what’s happening there?

We were recently one of seven winners worldwide for MIT Solve’s AI assessment challenge, backed by the Gates Foundation. The irony is that Frankenstories was designed to sidestep AI by instead focusing on peer feedback. However, we did that so well that students started producing a lot of writing and then teachers wanted some kind of bird’s-eye view so they could see how students were going and decide where to focus. Our pitch to MIT was technically the simplest in the cohort; we won the grant because the judges wanted to see AI feedback in a context that was all about creativity, collaboration, and fun.

TE: What are the most important ways Frankenstories stands out from other programs aimed at lifting students’ writing outcomes?

We are different in fundamental ways. First, Frankenstories creates a unique hivemind experience that fundamentally reshapes student attitudes towards creativity and collaboration. Second, flexibility. Frankenstories is a tool, not a program. You can use it in any subject, any grade, with any number of players from individuals to whole classes and beyond, and you can dovetail it with any writing program you happen to be using. Third, craft authenticity. Frankenstories is less like ‘doing school’ and more like doing real writing. Earlier I listed some of our main educational evidence bases, but there are also a range of ‘craft’ foundations that we draw on, not least of which is a rich body of knowledge derived from improvisational theatre and spontaneous creativity.

TE: Looking ahead, are there plans to refine/adapt the program as GenAI increasingly disrupts school education?

My friend John Warner has a great line, “ChatGPT can’t kill anything that’s worth preserving.” That’s my attitude. Frankenstories is about making AI models irrelevant by creating an experience that is inherently live, social, and human. Don’t get me wrong: I love the new GenAI models and use them all the time. But from an educational point of view, we need students to value their own thoughts, value the act of thinking, creating, & exploring, and want to get better at modelling the world through language so that they will be proficient and reflective users of whatever these new tools turn out to be.