Frankenstories

We're trying to solve the problem of access to higher-order writing skills.

Writing is a crucial skill in many domains, including personal, academic, and professional. It's basically how we model the world through language.

The problem is that writing becomes increasingly difficult to teach past Grade 4, when students begin to produce more complex texts, such as entertaining stories, detailed analyses, or compelling arguments. 

Educational researchers know what works—reading rich texts, mentor text modelling, joint construction, volume of practice—but doing what works requires class time and student attention, both of which are in increasingly short supply now that we all live in a 24/7 short-form multimedia casino.

Quantitative data about writing ability is hard to get. Globally, PISA doesn't measure writing. In the U.S. NAEP didn't publish their last assessment, so current data is over 10 years old. In Australia, we assess students nationally every year (though individual students are only assessed every 2 years). The Australian Educational Research Organisation published an analysis that found a steady decline in writing ability from Grade 3 onwards, with 85% of Grade 9 students writing at or below Grade 7 level. The last extant NAEP report, published in 2011, found a similar trend in the U.S., with only about 25% of Grade 8 and Grade 12 students proficient or higher.

The point is that we don't actually have a lot of data, but the quantitative data we do have (and certainly anecdotal data from teachers) shows there is a problem—and the problem is likely to be universal because it is related to a skill that is inherently complex.

(Possibly LLMs will render this problem moot, but unless we expect a future where human influence is irrelevant, we believe at the very least that students will need to be able to function as executive directors or collaborators with models, which still requires a base level of proficiency in the cognitive demands of writing.)

Frankenstories is a fast-paced, multiplayer, online creative writing game designed for Grades 5-12. 

It's designed to make writing so fun that students don’t want to stop.

Mechanically, the game is like a cross between Gartic Phone and Theatresports:

• Players play in groups of any size: individual, pairs, small groups, whole class. 
• Everyone gets the same prompt, then players write in timed rounds. 
• When the timer runs out, players read anonymised submissions and vote for their favorite. 
• The winner gets stitched into the spine of the story, and the process repeats.  

That simple format creates an engaging, social, hivemind-like writing experience that is inviting to even the most reluctant writers.

The game builds initial enthusiasm and stamina for writing that teachers can channel then into structured learning, either by using our library of skill-based curriculum prompts or by creating custom games tailored to their students' needs.

Also, while narrative is often the entry point for Frankenstories, it can be used for any kind of writing (analytical, informational, persuasive, poetic) and in any subject: ELA, social studies, history, science, foreign languages, you name it.

Here's a promo video to give you a sense of how it works.

Who is served by Microsoft Word or Google Docs? Everyone who needs to write.

Similarly, Frankenstories serves everyone who needs to *learn* to write.

We don't mean that glibly. A word processor is a blank page with some additional affordances; Frankenstories is a blank page with rounds, timers, customisable prompts, other players, etc.

So our learner audience is everyone who is making the transition from basic literacy to writing more complex texts.

This audience is identified by fluency rather than age, gender, location, ethnicity, class, etc.

Because learning to write is a complex, multi-year process, our span of learners ranges from writers who are only capable of fragments to writers who are proficient across multiple genres. (In grade terms, we have students from Grades 3-12 all playing the same underlying game.)

Also, because it's a blank page, Frankenstories is not restricted to English language learners. Although the interface is not yet localized, we've had games played in French, Malay, Mandarin, Korean, Portuguese, German, Hebrew, and more.

However, from a product/market perspective, our real audience is teachers.

While Frankenstories can be used by anyone—individuals, homeschoolers, groups of friends—we designed it with classroom teachers in mind.

We believe it is useful to any classroom teacher, but it is particularly useful for teachers of disengaged or low-ability students, for whom it can be hard even to write a few words.

Frankenstories can't work miracles, but it does give teachers a vehicle to get students engaged in writing. We then provide instruction and resources to help teachers build on that momentum and help students build increasingly complex skills.

Our teacher audience is really only limited by access to laptops/tablets and the freedom to experiment with our resources, though these are significant barriers in their own way.

Our co-founders combine three critical disciplines: classroom teaching, writing, and digital learning product design (including game design). 

Because we are not in a classroom, we talk to teachers using Frankenstories on a weekly basis and respond to feedback with new features, content, or product refinements. (We have received written or verbal feedback from more than 150 teachers across four countries.)

Frankenstories itself emerged from our prior work on our mentor text modelling resource, Writelike.org, which revealed a need for an engagement engine in classroom writing—so we understand the problem intimately, and have years of prior experience working on other solutions.

We released our first prototype in late 2021, which demonstrated the core game loop and convinced us that Frankenstories was a powerful engine for learning.

Since then, we have continually released new features and content, including teacher controls, approval modes, prompt libraries, support materials, engagement features, management and reporting features, subscription model, and so on.

• Over 2022 we grew to about 2,000 students/month who wrote 2 million words.
• Over 2023 we grew to about 10,000 students/month who wrote 7 million words.
• As of April 2024, we are at around 24,000 students/month who have written over 5 million words (2.5 million in March alone).

Approx. 20% of our users are in Australia/NZ, 80% are in U.S. and Canada. We don't currently promote in any other regions, but we have a small number of users scattered around the rest of the world.

We raised a seed funding round at the end of 2022.

Developing edtech products that genuinely make a difference is really hard:

• You need to innovate on content, pedagogy, and product simultaneously.
• You need continual feedback.
• You are limited by school calendars and curriculum plans.
• Teachers have minimal bandwidth for experimentation.

We've crossed several danger zones already. 

We've developed a product with high usage, rave reviews, and strong retention.

Now we are thinking about strategic partnerships that help us scale and accelerate growth.

To that end, we are interested in three things:

• Research partnerships: Obviously we'd love help getting some learner growth data, but we are also interested in ways in which Frankenstories can be used as a kind of laboratory for running general educational research experiments. We're interested in meeting researchers studying not only writing but also creativity, collaboration, and SEL. We think that would produce interesting new knowledge in its own right, as well as build the legitimacy of the underlying product. 
• PR: We could use help spreading the word.
• Distribution: So far we have been advertising to teachers on social media who then pass it along through word of mouth. We can keep growing that way, but we are starting to feel the need to approach things at a more institutional level. That could mean help with district contacts in the US, after-school or extracurricular organisations, international educational organisations—we're pretty open. At the end of the day, we see this as a diffusion of innovation issue.

Frankenstories turns writing into a team sport. It's not just that students are writing together, but that the writing process is broken into rounds with pressure timers and peer voting, and you can have teams 'competing' on versions of the same prompt.

This approach is completely new. 

Other products such as Storium and Frantic Fanfic are great in their own ways but don't have the same sense of compression and speed that creates the hivemind effect you get in Frankenstories. 

And none have the multi-genre curriculum scaffold that gives teachers a skill development pathway in different genres such as narrative, argument, persuasive writing, etc, as well as applying to other subjects.

Its closest analogue is Theatresports, but for writing instead of stage performance.

The effect is so striking that we wouldn't be surprised if the Frankenstories format simply becomes part of the way that middle schoolers are taught to write, whether through Frankenstories specifically or as a generic approach that others copy in their own implementations.

The point of Frankenstories is to get middle school students so enthusiastic about writing that they will willingly invest effort in improving their skills.

If we do that, we can get more students across the Grade 5-9 literacy chasm to where they see value in complex texts, learn from models, take feedback, and are largely self-sufficient.

Here's our logic model from the ground up:

1. THE CORE LOOP

The problem with most educational games is that the gameplay doesn't actually teach the skill. But in Frankenstories, the core game loop models the same skills we want to teach writers:

1. Write
2. Read
3. Analyze
4. Adapt

In-game time isn't wasted. In fact, an average game has that core loop repeated 5 times over 30 minutes.

2. THE ENGAGEMENT LAYER

That loop is useless if students don't want to do it over and over again. 

Frankenstories creates engagement through three levers:

• Surprise: Every 90 seconds, something new and unexpected is introduced and students have to adapt and respond.
• Pressure: Not too much, not too little, controllable by the teacher. Every phase has a countdown and a word limit and personal pride is at stake.
• Pleasure: The craft pleasure of reading and writing good entries, plus the social pleasure of competing and collaborating in a community.

3. THE LEARNING LAYER

That's fun, but how do students develop skills?

• Cognitive load: Everything is broken into small chunks, there's lots of repetition, and students learn by modelling the winners.
• Social learning: Students are learning from each other while engaged in a meaningful, productive activity.
• Genre pedagogy: Students are writing authentic texts for a valued audience (peers), learning genre patterns across multiple text types.

This learning layer is supported by a library of over 800 skill-based game prompts that can take players from jotting down observations to constructing complete narratives, arguments, etc.

And that library is itself supplemented by Writelike.org, which has mentor-text modelling activities matched to most of the game prompts.

4. DEVELOPING GENERAL CAPABILITIES

This framework doesn't only teach writing skills, it also develops two general capabilities:

• Creativity: Observation, free association, combinatorial thinking, courage, risk, embedded in a practical craft context.
• Collaboration: Selflessness, attending, yielding, responding, contributing, creating a culture of psychological safety.

5. PROVIDING TEACHER SUPPORT

From a teaching perspective, Frankenstories is a typical "minute to learn, lifetime to master" tool. We provide extensive PD resources to support teachers with content & pedagogy.

EFFECTS

Following this model, students develop greater interest and enthusiasm for writing and increase their stamina. 

If a teacher uses Frankenstories skillfully, targeting their prompts and scaffolding, then students also develop more advanced writing skills.

Better writing skills lead to increased educational achievement, better behaviour, greater professional and college readiness.

CAVEATS

• Students need to also read more advanced texts (but Frankenstories can increase motivation for reading).
• Frankenstories takes time (but learning to write takes time—and Frankenstories uses time efficiently).

BIG PICTURE, LONG TERM

The policy-level measure is proportion of students who are at grade-level or above in writing tests. 

Since only a few jurisdictions test writing as an isolated skill, that really means grade-level performance in subject tests with a heavy writing component, e.g. ELA, history, social studies, science.

More broadly, this is also about developing a pro-school affect, so we would hope to influence increases in school attendance, completion, reduction in behavioural incidents, and so on.

But we're a long way from that scale, and drawing a throughline from Frankenstories to these results will be challenging.

HERE & NOW

To date, we've relied on teacher feedback to give us a subjective sense of student engagement:

• "It's truly, and I don't exaggerate this, it's truly the best thing that's engaged kids in writing." —4th Grade, U.S.A
• "Definitely engaged my disengaged writers, that's for sure." —9th Grade, Canada
• "Such an amazing resource! The collective yelp of indignation when the timer runs out (by recalcitrant writers who normally refuse direction to write a paragraph) is absolutely priceless!!" —9th Grade, Australia

Plus anecdotal feedback about skill development:

• "They did the Settings lesson on Writelike and then played the game and their writing improved immediately." —8th Grade, Canada

• "We think it is valuable for them to see other students' examples of writing...it has made many of them step it up because they see what others are capable of, and they want to impress others, so they do their best." —6th Grade, Australia

• "As we've used it more and refined the instructions they know to be looking for what is said in the instructions in the responses of their peers.  It's become a peer feedback mechanism as well as a writing practice tool." —4th Grade, New Zealand

NEAR FUTURE

We're currently prototyping ways to evaluate student progress over time.

A key feature of Writelike and Frankenstories is that they both focus on the kind of writing—short form, context-dependent, highly variable criteria—that has historically confounded automated feedback.

GPT3.5 crossed the threshold. We didn't rush straight into implementation, but we've been doing testing and evaluation and now have plans for automated reporting for teachers, which will display as a longitudinal heatmap evaluating dimensions such as:

• Word count: Turns out to be an accurate proxy indicator for quality with beginning students. 
• Content: We can use an LLM to assess content quality (e.g. amount of detail, relevance to context).
• Coherence: We can use an LLM to assess readability (e.g. punctuation, run-on sentences).
• Voting quality: In Frankenstories, voting is as important as writing. We can use any/all of the above indicators to track voting quality.

Once we build this reporting layer, we will be able to quantify some of Frankenstories' effects and provide more helpful guidance to teachers on how to use the tool.

Typical web app stack: Azure, .NET, React.

We have plans to use foundation LLMs for evaluation and reporting, as described above. These will be standard models, used as a service.

On the content side, we use diffusion models such as DALLE and Midjourney for image prompts. The best part of that is making bespoke images for student groups in regional locations who don't typically see themselves represented in middle school adventure imagery.

4 full time

2 part time

2.5 years

We're still a small team who work together because we've all known each other for 10 years. 

We're a 50/50 gender split.

We're all-but-one Australian anglo, so not especially racially diverse. That's more a product of chance than choice. 

We're clinically neurodiverse.

Internally, we have a flat, collegiate structure. High transparency & openness. No egos. Everyone contributes. Lots of room for agency.

We basically enact a set of collaborative values on a daily basis: selflessness, truth, detail, diversity, craft.

We organise work in a way that maintains pace but doesn't burn anyone out and responds to individual needs and circumstances (parenting, caregiving, etc).

We treat everyone on the team equally. Everyone has a mix of interesting work and grunt work.

We're highly solicitous of feedback from teachers of underprivileged students. Part of our mission is to help marginalised populations develop skills and cultural capital. We're still just building a beachhead there.

Possibly a distinctive feature of our team is we're very conscious of the class-stratifying effects of language and writing in general. We try to decouple quality from register, as it were—the whole Shakespearean high and low—encouraging students (and teachers) to write all kinds of content but do it with a sense of craft and skill.

We have a freemium SaaS model within a standard company structure.

Our reasoning is as follows:

• We want to be self-sustaining and not reliant on grants, which come and go (and are not as available in Australia as in the U.S.).
• We're a small team that can operate quite efficiently. We don't need a huge amount of revenue.
• We don't want the compliance overhead of running a not-for-profit.
• We want to provide some kind of free experience that anyone can access.
• We want pricing to be as close as possible to a no-brainer and affordable for individual teachers so they don't have to try to convince their admin to pay.

Which leads us to:

• Teachers can play casual Frankenstories games for free. These games can have any number of students and the teacher can customize the prompt text and image to an extent.
• We have a Pro subscription, which is US$60 per teacher per year, with unlimited classes and students, which provides more customization, content, and class management features.

We don't know if this is the best model, but it's the one that makes the most sense to us.

As described above, we have a freemium subscription model that is based on per-teacher pricing. 

We don't know if our business model is optimal. We find that one teacher will discover Frankenstories, try it, share it internally, and then staff will lobby for a centralised purchase, which means budget allocation, which can mean a one-year wait to earn revenue.

We are steadily growing revenue with each school year commencement (northern and southern hemisphere). To reduce the barrier to entry, we offer a generous discount for new school year purchases.

However, cycle times are quite long, which is why we think we need to start engaging with district offices, distribution partners, and other institutional stakeholders.

We've been privately funded to date, and our investors are actively involved in the company.

Based on our strong relationship with these local angel investors, we have confidence in our ability to raise capital as needed to maintain operations as we build our subscriber base.