Automation & Design Sprints
Design sprints are a lot like particle accelerators. You get a handful of interesting profiles in a room, throw in some trends or insights from your research, and hope to get your mind blown. Or at least that is the intent. However as many might recognize, the process is often closer to an arduous hurdling race than a frictionless whooshing of subatomic particles.
What remains true across both analogies is the fact that design sprints require large quantities of energy, a safe space for explosive (creative?) effervescence and a keen eye for nuanced and sometimes seemingly imperceptible insights.
Over the past year or so I have been experimenting with new workshop activities that involve very simple, openly-accessible technologies to try and make workshops an enjoyable environment for reflection through iterative thinking. The experiments are interactive concept generators that throw features and user needs together to create new, unexpected, solutions for participants to reflect on and tweak to their needs.
I’ve noticed design sprints are often inspiring experiences for participants but the quality of the design solutions that emerge from them greatly varies. Most ideas will either be low-hanging fruits, unrealistic, or aspiring vapor-ware: lacking substance or too loosely rooted in the given context.
The quality of an idea is always challenging to critique since it can seem very subjective. There are many tools and activities like prioritization matrices and defining design goals used to ensure participants are working towards a shared vision with a cohesive intent. Nevertheless underwhelming flat ideas persist defying all rules of common sense and avoiding the lush playful pastures of a fruitful imagination.
So why do workshop ideas often fall so flat? One possible reason is the lack of room for critical thinking or simply critique in most design sprints. Giving and receiving feedback is a core part of the design process as wonderfully highlighted by Pentagram’s Natasha Jen in her 99U talk on the importance of the crit as a way to regularly compare design solution and the check the evidence supporting it. She goes on to call out the limitations of packaged design thinking sprints in their inability to do justice to the iterative nature of design. To allow for iterative design, participants must not only have time and readily available evidence of the realities they are designing for (both equally tall orders), but must also be willing to have nuanced explorations of a topic, discussing it, unpacking it.
One of the safest go-tos to encourage participants to relish in the joys of nuanced ideation are sacrificial concepts. They can be short storyboards, one liners, or easy to scramble card decks that give concise yet incomplete concepts for the participants to react to and iterate over.
Over the past year or two I’ve been slowly exploring how chance-based activities can help improve the experience of using such sacrificial concepts. By chance-based activities I mean looking at ways of randomizing contribution from ideation sessions and scrambling them together to create new scenarios.
The main assumption from my part being that by making it easier to incrementally tweak ideas, I might free up some mental space for participants (and save time as all good tech must promise to do) and foster deeper conversations around the intricacies of the concepts being brainstormed. Stemming from a broader history of generative tools being used in research contexts to facilitate communication, I simply tweaked existing formats that have helped me in my practice.
Below you will find a series of experiments drawing from this space; these are still works in progress. Even this post is intended more for personal archival purposes than anything else. In this piece you will find:
- More detailed thoughts on the challenges of translating insights into actionable design solutions
- How my own work on the topic at Quicksand has evolved
- A collection of works that have informed my experiments
Take away what you may!
1| Notes on hairy ideation sessions
The observations that follow are incomplete; personal notes on limitations that might be encountered when running ideation sprints or participating in them.
Many sprints start with a round of sense-making, reviewing quotes from research that would have preceded the workshop. Whether doing it as a member of the research team or as a workshop participant, information overload is hard to avoid. You are presented with user needs, user constraints, macro and micro insights. There are a slew of informational nuggets to absorb.
Forming a comprehensive picture of the topic, let alone starting to tackle the topic can be a challenge in and of itself: Where to start, what to prioritize, how to do justice to the needs and realities presented?
Two challenges are intrinsically linked here: the volume of information and the complexity that intricately threads each byte of information to the ones surrounding it. I have yet to find tools that properly bring together the systems angle of exercises like problem trees or stakeholder mapping (or ERAFs) with the playfulness and lateral thinking of ideation.
Returning to the particle accelerator analogy, combing through user contributions in search of “aha moments” is a little like carefully rearranging subatomic particles in the hope of uncovering secrets with disruptive interplanetary repercussions: it’s a tedious task powered by the hubris of wanting to solve universally relevant challenges.
This process of sifting through large amounts of gripping and intrinsically human stories leaves many either vulnerable to the temptation of oversimplification, a false sense of objectivity and an unhealthy savior complex or on the other hand, a hollow feeling of helplessness in front of towering structural issues that can’t be properly addressed through the service or product being conceived.
The most common remedy for this malaise, a cornerstone of packaged design sprints are How Might We statements.
“[HMW statements are] often used for looking at insights gathered from research and framing them into opportunities or alternatives. When teams are feeling stuck, HMW can be used to reframe the current problem and move the team forward.” Mozilla’s Open Innovation Toolkit
See for yourself with these few examples from recent Quicksand projects:
How might we empower micro-entrepreneurs to grow their businesses?
How might we provide appealing alternatives to non-renewable energy?
How might we enable self help groups (SHGs) to provide accurate ASRH information to the homes of their members?
The beauty of HMWs is their versatility, their adaptability to verbally wrangle any challenge. They are also the first step in uprooting an insight from its context; a dangerous but necessary task. If not done properly, designers, workshop participants, and researchers alike will find themselves aspiring to end world hunger by the end of the sprint. IDEO also calls attention to this point of contention in their Design Kit:
“Finally, make sure that your How Might We’s aren’t too broad. It’s a tricky process but a good How Might We should give you both a narrow enough frame to let you know where to start your Brainstorm, but also enough breadth to give you room to explore wild ideas.”
The popularity of HMW statements points to a clear need for concisely capturing opportunities when concluding a research download in order to build consensus on the opportunities they represent. How can this format be improved to avoid participants going too broad while encouraging complex inter-sectional viewpoints?
The other challenge in design sprints is the sprinting itself. The notion of sprinting implies a flash performance, a short burst of energy with a straight-forward end goal: crossing the finish line first.
For one, this mindset is not necessarily conducive to reflection, but more importantly it glorifies a certain kind of creative performance: to be the bold, dashing, witty designer with all the cool disruptive ideas. There is also the performance of the sprint organizer and their need to ensure a certain succession of recognizably (pre-packaged?) creative activities to their client who also happen to be participants. I’m not sure if this counts as conflict of interest but it definitely is shooting yourself in the foot…
My question is how can ideation be more about contributing to a collective and constructive sense-making effort rather than a performance of pseudo-creativity? Ideation should be about opening doors to new perspectives, growing appetites for curiosity and imagination, not obsessive solutionism.
From what I’ve seen in our sessions, it can be difficult to have deep discussions on the specifics of a given challenge, to formulate solutions not to stop at them but to further reflect on the given topic. This is where sacrificial concepts are great tools to explore the tradeoffs, risks, opportunities, feasibility, desirability of an idea and build on them. Each new concept can then be collectively assessed and critiqued (an experience made even more fruitful when done in the presence of end users, field practitioners).
But I find that the use of sacrificial concepts is antithetical to the notion of sprinting. This is where the friction comes in: the act of repeatedly refining and unpacking an idea can be an exhausting task. This is where the hurdling race seems like a better analogy, albeit less glamorous, than the sprint. Iterating on ideas, knowing what to keep and what to let go of is a time-consuming and sometimes repetitive process. It can be hard to keep participants engaged if their involvement with the project at hand is limited to the duration of the workshop. There is a slightly different format of communal creative production (which also draws from the world of terrestrial locomotion sports) the marathon. Its many derivatives: makeathon, hackathon, gameathon, have successfully embraced the idea of the grueling accomplishment that requires a sustained effort.
I wonder what makes these formats worth the sacrifice? Is it just a different audience and interests? The joy of braving the seemingly impossible? The stake each participant carries in the event? These events’ focus on prototyping and making? The playful competitiveness they foster?
The creati-rative fatigue that emerges from both design sprints or makeathons (of ideating, iterating, prototyping) is similar in many ways. My sense however is that one is better at building on participants’ strengths and making the stakes feel real. The nature of a makeathon also nurtures a sense of ownership that comes from crafting something tangible. I don’t think the experiments I will present below necessarily check all these boxes but they will try to help participants crystallize their intents and tweak them as one might a cardboard prototype.
2 | The experiments
Each one of the tools, games and activities referenced above lives at the intersection of chance operation, human interpretation, and collaboration. I really think using these playful activities with a dash of technology can help make workshops a lot more fruitful. This would however require these tech-enabled formats are as accessible and adaptable for workshop organizers as creating paper templates prior to a workshop.
In an attempt to make tweakable, chance-based, idea generators, I decided to create idea scramblers in Processing. Part automated POV statement and part Cosmocracy format, the idea scrambler uses templatized concept statements generated in a workshop to create new sacrificial concepts.
First used at an ideation workshop on the future of contraception where a range of stakeholders, designers, chemists, field care givers, investors and project managers were asked to better understand how the contraceptive technologies they were working on would affect (positively or otherwise) the end users’ day to day. Part of the workshop consisted of formulating concepts by using the following statement:
“A [qualifier] [contraceptive method] to help [intended user] [objective]”.
It was first provided to the participant groups as a physical madlib template for them to fill out. The suggestions were then added to an Excel linked to the program to generate new, playful and sometimes creepy solutions. Nevertheless, this process successfully married the machine’s effortless ability to form new sentence permutations and the participants’ ability to react, reflect and weave them into new provocations or stories relevant to their discussions.
The experiment validated my hunch that there was value in scrambling strategic components of ideas instead of seeking drastically new ones constantly. This approach made it much easier to see the potential of iterative changes and encouraged the participants to create several variations for each sacrificial concept. After using several versions of the idea generator on recent projects, I’ve learnt that there is clear value in a more permutative approach to ideation. Seeing components a human mind wouldn’t pair together are placed into a sentence challenges the viewer to make sense of it leading to new, fresh ideas.
The final version was made more ideation-friendly with an interface that could allow users to only scramble certain parts of the concept making it easier to iterate upon. For instance, you can select one end user say, a doctor working in a rural area your group is designing for, and then have the rest of the statement continue scrambling thus only keeping only the end user the same. The inspiration for this was Manetta Berrends’ Cyber/technofeminist cross-readings, a search engine that parses through a given set of manifestos for any given single keyword and nicely curate the responses.
That being said, technology for technology’s sake never helped anyone. This scrambling process turned out to be as valuable as an analog, post-it-based, scrambling design method as much as it was as a digital generator. Neither did it guarantee further debate or discussion without external prompts. It is yet another tool amidst a sea of options that will require further fine-tuning and applications to shape its own purpose. It has nonetheless helped crystallize the value of a permutative and generative approach to ideation in my practice which I hope to explore further.
For now you can find the latest version of this scrambler (as in the GIF above) here. This approach is finding its way into other Quicksand projects. Here are a few examples:
One of the simplest, playful, and public applications of the format has been for the branding of Quicksand’s latest Unbox Festival, a three day event that brings together the global community we have developed over the years. The generator was used to highlight the festival’s open unconference format that allowed for chance encounters and rich discussions. Each time the page would load, the scrambler would put two different events happening at Unbox together and assign them an interaction hinting at the spirit of alchemy that drives Unbox events.
We also used the scrambler ina project on the future of humanitarian aid commissioned by the Barbican’s Life Re-Wired Hub in London. There it was used throughout the course of the design process both internally and with groups of design students to explore what a decentralized response to a water crisis in India might look like. The idea here was to rephrase the statement to help ideators focus on self-organized groups and citizens as key actors in disaster response, the scrambler in a way planted seeds for longer stories or journeys to be depicted in our final exhibit.
The generator was both used as an activity during the workshops with the same paper-based fill in the blank activity and then was translated into a storytelling device on the project’s the website and the exhibit itself, to emphasize the multiplicity of possible futures and realities that had been explored during the brainstorms.
A very interesting offshoot of these experiments was to collaborate with my colleague Salil Parekh to put together a workshop that would uncover the secret sauce that makes conspiracy theories so so tasty whether you believe them or not.
For this the usual scrambler wasn’t enough. We wanted this workshop to be about the future of misinformation and use some of the tools that might be used to invent news in a not so distant tomorrow, so we needed a much more inventive generator.
Instead of the usual scrambler, we had participants review headlines generated by a machine learning algorithm Salil had trained on 2018 headlines and then choose one of them to investigate further.
The investigation was conducted through an activity inspired by games of Wikipedia page hopping (another semi-chance-driven format) where participants had to find a Wikipedia article related to hints found in the machine-generated headline and then weave together a conspiracy by collating evidence from the ten successive links they used to hop from one page to another (granted this process would also be fun to automate…).
Here are Salil’s notes on his process with machine learning:
“Two processes were used to generate content using machine learning. The fake news headlines were generated using Tensorflow, a high level machine learning language developed by Google. Although this isn’t easily accessible or easy to use, it is relatively very simple for those who regularly use machine learning algorithms. This machine learning model is called LSTM (Long Short Term Memory), and long story short, it works in a similar fashion to the text prediction functionality on your smartphone, predicting which character/letter will come next. To generate the fake news, I first had to train it on some real news, so I had to get a whole bunch of news headlines covering a wide variety of topics to ensure lots of diversity in the output. It’s easier said than done, but a custom python script was written, which worked all night long, ripping all the headlines from 2018, from a popular Indian news website. Now that the training data had been collected, the machine learning model needed to train on this data in order to successfully generate the fake headlines. After another few nights of training, we finally had some very interesting results! Not only were the outputs very diverse, but the ML model had also generated new words, which was very nice to see.
We used another ML algorithm to then create images from these headlines. This is a lot more complex, so we used Runway ML, a brand new software which makes using ML models a breeze thanks to its visual interface. The model used for this task was the AttnGAN model, and although it doesn’t generate photo-realistic images, or any sort of recognizable objects, the images were exactly what we were looking for. The low resolution, blurry images had vague forms which left a lot to imagination–exactly what conspiracy theorists would use! The imperfections in both the LSTM model, which lead to the creation of new words, and the blurry images from the AttnGAN model were very handy for our intents and purposes.”
Future of Consent Stories
As might be evident in the examples above, be it as a scrambler or ML headline creator, these experiments have essentially been limited to outputting design briefs, short provocations, similar to How Might We statements. They bring nuance to the discussion when compared to one another but not necessarily when used in isolation.
To build on this observation, we facilitated (with another colleague of mine, Sahil Tandon) a session around a new game that explored the future of consent and the variety of forms it might take in the future. Starting with mind mapping and good old madlib exercises, the sentence templates here were different from one another to allow participants to look at an issue from different perspectives:
“How might [subject] convey meaningful consent when [action / situation] by [agent]?”
“How might a[agent] assess meaningful consent from [subject] when [action / situation]?”
“What does consent look like for [subject] when [action / situation]?”
These questions were then expanded into a series of situations and choices that formed convoluted epics when woven together resulting in a collaboratively created “Choose Your Own Adventure game.” A viewer could come up to the web of choices and explore the ambiguous situations in which their understanding of consent might be challenged in the future.
This is a very exciting format that I am hoping to continue developing both in analog and digital form. We’ve tried some experiements using the open source software Inky, designed specifically by Inkle to create interactive narratives.
Social Media Generator: So far all the generators (apart from the machine learning version) have used predefined words or phrases that they scrambled. Ideally these generators could draw from unique words found online on social media or Wikipedia to create even more surprising results. On top of that, pushing the machine learning front with ml5.js or basic language processing libraries could make this format much richer.
Visual concepts: Visualization through illustrations or photo collages can go a long way to help crystallize a sacrificial concept. I have been looking at ways to generate visuals through ML or hand-drawn images strategically laid out to quickly create snapshots of potential ideas.
Trouble Shooting: The big missing piece is the idea of debate and helping participants feel comfortable exploring risks and tradeoffs their ideas might engender. I’d love to make a cautious idea generator or solution complicator that would take a solution and tell stories of its misuse (in other words a Black Mirror plot generator).
Developing the human side: It’s taken me a while to get the program to a state that can be shared so updates might have to wait a bit. I’d like to now focus on how this can be used (by humans) in as many conversations and workshops and build on more elusive values of design thinking like the bridging of empathy at an individual level while factoring in societal and infrastructural systems, a point well captured by John Payne his piece on the place of empathy in design.
3| Existing references
When collecting these references, I was looking for a few different things:
- Easily generating large amounts of ideas
- Constructively identifying and leveraging constraints and responding to user needs
- Capturing ideas in ways that allow for comparison, debate and tweaking
- Allowing more room for improvisation and play
A great place to look for inspiration and partial answers has been generative and co-creation tools. These engaging activities are aimed at making co-ideators feel more comfortable and inventive with the topic at hand.
Models of Impact: A workshoping game created by Matthew Manos. As one of the very first instances of generative ideation I’ve come across it continues to be one of the most complete and constructive formats I’ve come across yet.
Conditional design: A group of colleagues Luna Maurer, Jonathan Puckey, Roel Wouters and Edo Paulus who invented a collection of visual games predicated on the definition of visual conditions for participants to collaborate around.
Thing from the future: A card game created by Stuart Candy and James Watson that has players compete and collaborate to imagine objects of the future according to a specific set of prompts.
Cosmocracy: A board game that uses randomly drawn cards to create laws to be debated in the cosmic congress.
POV madlib: A common exercise in workshops used to, like HWM statements, reframe challenges into actionable (and relatable?) statements. In their description, Stanford’s D.School, recommends using the POV madlib to then inspire HMW statements. I could not verify the origin of the tool hence the link to Stanford’s D.School one of its earlier advocates.
Creative.ai: A studio working exclusively on the use of human and machine collaboration in creative industries with an amazing list of generative design works and references here.
At the end of the day, much of a workshop depends on the humans present in the room and the way workshops are structured. I really think the particle collider analogy still holds in that the key to avoiding the same old tired ideas might be in enabling participants to experience ideation as a form of experiment at the intersection of particle physics and alchemy. That ideas don’t have to be breakthroughs as much as they are journeys, that ideas do not spawn and fade over the course of a design sprint. These ideas will evolve, grow and have lives of their own in each participant’s mind contributing to the creation of new shared imaginaries. So far this is what the idea scrambler has most successfully represented, how ideas can have lives of their own, outside of our busy schedules and project deadlines.
These experiments are still a bit basic for now but being able to make them approachable and adaptable by non-programmers like myself has been a very time consuming endeavor that required trying out various ways of integrating it within our workflow at Quicksand. Thankfully open source software like Processing that have a wealth of documentation online and are designed to be easy to pick up provide a great opportunity to create custom facilitation tools for designers.
Please do reach out with suggestions, inputs or tools of your own. You can find the concept generator here if you missed the link in the text above.