Brainstorming is the go-to method for all creative thinkers. Few other methods can support idea generation and promote innovation so effectively in such a short space of time. So brainstorming works - but is it optimized to the extreme?
We asked ourselves this question at the start of the project. Our own experience in design studies has shown that brainstorming is far from perfect.
From annoying post-it notes to influencing votes: there are pain points in brainstorming that make it difficult to generate ideas efficiently and without stress. In order to not only think from a designer's perspective, we started a detailed pain point analysis in our network before concept development with the simple question: What are the most typical problems that occur during brainstorming?
The sheer volume of responses confirmed this: Yes, brainstorming has problems. And not too few of them. It was striking that some problems were mentioned over and over again - mainly difficulties in the structure of brainstorming and thus also the lack of discipline to adhere to this structure.
”Hey creatives - yes, that means all of you! A quick one-question survey for you: When it comes to brainstorming new ideas - be it in groups or by yourself - what is your biggest issue? Timekeeping? Getting off topic? Agreeing on a ruleset? Collecting results? Just drop me a keyword, a short sentence, a drawing in the comments…. whatever comes to mind.“
Before brainstorming can be digitized, it is logical to first understand how analogue brainstorming works. Early user testing was the ideal way to do this. This resulted in fundamental findings: in particular, a clear picture of the structure.
That structure: Think, Sort, Vote. We’ll get to that later.
User testing was used to answer various questions: How is brainstorming currently used? How is a voice assistant accepted in this context? Can the test validate our idea? Which voice commands are used intuitively when none are given?
Two groups of 3 test subjects were asked to solve a problem via brainstorming. The first group was provided with the regular utensils of a brainstorming workshop, i.e. post-its and pens.
Timekeeping and moderation were handled by a voice assistant that we simulated. We simulated all the functions from an observation room and made them available to the brainstormers via a loudspeaker and projector.
Group 2 was the control group and worked on the same task. However, the normal utensils were missing here - so they were forced to brainstorm entirely with our assistant. No voice commands were deliberately given - to find out which questions and commands are used intuitively.
Based on our experience, the survey and early user testing, we prioritized the requirements that form the essence of our prototype. Or: requirements that our tool must fulfill.
To summarize …
Just like analog brainstorming, VIA is divided into three stages:
Think, where ideas are generated.
Sort, in which these ideas are sorted.
Vote, where the ideas are voted on.
Each of the three phases is the optimized version of the analogue equivalent. The principle and structure of the actual brainstorming is adopted—but the analyzed pain points are eliminated and improved.
Keeping track of time. VIA takes on the task of keeping an eye on the time. Timers are set simply by voice and the countdown appears at the top right of the screen.
VIA takes notes for you. So that the focus is on generating ideas and a lively discussion can develop, VIA writes in the background and displays the ideas mentioned in individual cards on the screen.
“Why don't you try...” If it remains quiet for a while and no new ideas are mentioned (or a method is triggered by a request), VIA suggests a suitable brainstorming method. This provides a new perspective and re-initiates the generation of ideas.
“Have your say, too!” If VIA detects that the share of the conversation is too unevenly distributed within the team, it displays an unobtrusive notification. The participants who say little are thus asked to share their ideas without being judged.
Summarize ideas. In analog brainstorming, the ideas are sorted and grouped in this phase. VIA can form and name clusters using voice commands. All ideas are numbered so that they can be addressed quickly and easily. For simple interactions - such as moving a single idea - it is faster to do this manually. In interactive scenarios, such as on a smartboard, ideas can be moved by drag and drop.
Neutralizing the votes. A major problem with analog brainstorming is the pre-evaluation made by the first participant in the vote - the post-it that is voted for first tends to have a high chance of becoming one of the “finalists”. Voting shouldn't be this biased - that's why VIA doesn't show what the previous participants voted for. This means that the last points to be awarded are no longer tactical decisions.
Round and round we go. After the third stage, another round can be started with the best ideas. The process starts from the beginning, the only difference is the color of the ideas - each round has its own color for clarity.
Take the ideas with you. Every team is different - that's why importing and exporting is possible at any time. VIA suggests an export at the end of each round to keep the brainstorming within a suitable framework.
VIA makes itself known. In order to clearly display VIA’s current status, its logo is variable: different statuses are shown in different symbols.
Voice assistance is no longer uncharted territory - but detailed design rules are still missing. The course was able to establish initial rules for the design of voice, in particular the extent to which users want to be addressed and what expectations they have of their voice assistant.
The choice of words in the voice context must be adapted. During early user testing, we realized that addressing users directly via voice has a completely different effect than addressing them via GUI.
This can occur despite an identical choice of words in both interfaces - if you are addressed by a human voice, you feel attacked more quickly than with a written text. A good example of this is the reference to the proportion of speech.
In the user testing, the test person feels personally attacked, but as a notification in the finished interface, the offensive address and thus the provocation is missing.
Voice conversations are not linear. In contrast to GUIs (Graphical User Interface), VUIs (Voice User Interface) cannot rely on clear and linear communication.
While a simple button can be labeled “Next” in the GUI, all possible next functions must be programmed into the VUI. This means that in addition to “VIA, next!”, the user must also be able to say “VIA, next step”, “VIA, next!” and similar.
Ideally, a voice command should always work and thus come as close as possible to real communication - i.e. without “Sorry, I didn't understand that.”
Longer spoken texts overwhelm the user. Early user testing has also shown that longer texts - which are not displayed visually in the GUI - can be overwhelming.
The user is forced to listen carefully and cannot support listening by reading at the same time. The following situation is an example of this:
The test subjects instinctively reach for pen and paper to take notes, even though our instructions made it possible to ask them again - and provide a visual supplement - at any time.
Proactive interruptions are tricky. The voice assistant must always remain in the background - this hypothesis was clearly confirmed in user testing.The user always wants to remain in control, especially with an interface as human as voice.
If the assistant then switches on proactively without an interactive trigger (e.g. an expired timer), this is often perceived as interrupting and patronizing.
This effect can be mitigated if an acoustic or visual signal “forewarns” the user and prepares them for the upcoming interaction with the voice assistant.
Voice must be able to adapt to all contexts. Similar to GUIs, each scenario requires its own design - this also applies to voice assistance.
An example of this is the use of VIA in various situations: on a smartphone, for example in a café, VIA must not make any proactive announcements and should remain acoustically in the background.
In the meeting room at the smartboard, VIA has completely different requirements - here it can be “loud” and requires a different user approach.
Voice and screen must be clearly separated. A clear structure must be created and maintained during the design process.
Mixing different interface forms creates a chaos that ensures that the optimum display option is not selected.
It helped us to create a matrix (sorry, in German) in which input and output are clearly assigned to an interface:
And because there haven’t been enough videos on this page, you can watch me talk about Do’s and Dont’s in voice design here on a conference.