Skip to main content

Computational Stress in Production


Last night I attended MiniCAST, an online version of the Association for Software Testing's famous CAST conference. I've never been to CAST in person but I can say that the vibe here was great, much more informal and peer-based than the presenter-audience split I've seen elsewhere. It ran for four hours and squeezed in four talks on two tracks, several socialising sessions, and a keynote from Rachel Kibler.

Rachel spoke about stress cases, those scenarios when context, or the product, or both in tandem distress the user. For example, the health-tracking app that excluded women because it didn't include menstrual cycles, or the social media app that pushed a daughter's photo into a dad's timeline with a celebratory whoop ... on the anniversary of her death, or the ride-share app with numerous pop-ups that is hard to use in the dark, walking fast, with low battery, trying to get a lift out of a bad neighbourhood.

These kinds of threats to inclusivity, emotional stability, and personal security are seen in development process with low diversity, a focus on success, and a lack of interest in users and their real life situations. 

While not always common, stress cases should not be dismissed as simple edge cases (traditionally, a situation where some parameter is pushed to an extreme value). They affect real people in real, tangible, consequential ways. In our ROI-driven world this may not be enough of an argument for some software producers, but the potential for reputational damage probably is.

To help to avoid cases of stress in the wild, Rachel suggested a few approaches in development:

  • Have a designated dissenter, someone whose role is to look for the flaws, find the stress points, advocate for those who find themselves off the happy path.
  • Run pre-mortems, where the potential bad outcomes are written up as headlines and then routes to avoid them are found.
  • Read copy aloud in a bright voice. How does it sound when the content doesn't fit that medium?
  • Give some of your personas traumatic back history.
  • Put yourself under stress when testing. How does that feel? Where does the product fail (you)?
  • Be bold in telling management to be kind, considerate, and ethical.

Remember, there is no average user and someone is always having a bad day.


 Sarah Aslanifar talked about computational thinking which she described as:
an iterative system of generative reasoning in which people build models of a subject in a notation capable of being executed objectively and automatically be a machine, with observable and falsifiable output.
This style of thinking is the result of a logical progression from concrete to abstract thought through human history: oral, written, and now computational. As I understood it, at each stage it was possible for there to be dialogue at a greater remove from reality and at a greater distance between participants.

We're in the computational phase now and our abstractions, or models, have the potential to be encoded and executed. Monte Carlo simulation, where scenarios are run numerous times to understand the space of possible outcomes from some starting situation and with some set of constraints, might be an example.

I don't think Sarah said it explicitly, but the key thing here seems to be the use of the computer as a tool to aid thinking. Exercising a model independently of our own heads gives us a chance to reflect on where it is successful and where it deviates from reality. Analysis of the results can help us to determine what to alter to try to make it better.

Machine learning seems like an interesting area of this space. It is notoriously hard to interrogate, although it is certainly possible to experiment with parameters to improve its outcomes. A generate-and-test strategy is reasonable for exploring an unknown area, but it's not clear to me that it would qualify as computational thinking, not least because of the falsifiability requirement in Sarah's definition.

Perhaps I should have asked Alex Eftimiades about that. He presented on the challenges and value of testing machine learning systems in production. Production for him is financial systems, and the goal of his work is to inspect the firehose of data looking for potentially fraudulent transactions.


One of the points he made early on was that in the "traditional" software testing world, there is a culture of binary pass/fail decisions, where a fail typically indicates some kind of bug. In the machine learning world that sharp distinction is smooshed out into a probability distribution where uncertainty around a result is the norm.

Without a guillotine oracle, the approaches open to testers are to question performance and divergences. These are still comparisons, because testing is about finding differences that make a difference, but they are now statistical in nature. 

Without going into the technical weeds too much, Alex asked questions like does the performance of the system on its training and production data differ by an amount that is not explained by baseline variation? If I tweak the inputs to the system in known ways does the output of the system change in step in ways that are explainable and reasonable? Can I create a threshold for alerting by adjusting it until the balance of true and false positives is acceptable to me, in this context, at this time?

Comments

Popular posts from this blog

The Ideal Test Plan

A colleague pinged me the other day, asking about an "ideal test plan" and wondering whether I could suggest something. Not without a bit more information, I said. OK, they said. Who needs the plan, for what purpose? I asked. Their response: it's for internal use, to improve documentation, and provide a standard structure. We work in a medical context and have strict compliance requirements, so I wondered aloud whether the plan is needed for audit, or to show to customers? It's not, they replied, it's just for the team. Smiling now, I stopped asking questions and delivered the good news that I had what they were looking for. Yes? they asked, in anticipation. Naturally I paused for dramatic effect and to enhance the appearance of deep wisdom, before saying: the ideal plan is one that works for you. Which is great and all that, but not heavy on practical advice. --00-- I am currently running a project at the Association for Software Testing and there is a plan for

Notes on Testing Notes

Ben Dowen pinged me and others on Twitter last week , asking for "a nice concise resource to link to for a blog post - about taking good Testing notes." I didn't have one so I thought I'd write a few words on how I'm doing it at the moment for my work at Ada Health, alongside Ben. You may have read previously that I use a script to upload Markdown-based text files to Confluence . Here's the template that I start from: # Date + Title # Mission # Summary WIP! # Notes Then I fill out what I plan to do. The Mission can be as high or low level as I want it to be. Sometimes, if deeper context might be valuable I'll add a Background subsection to it. I don't fill in the Summary section until the end. It's a high-level overview of what I did, what I found, risks identified, value provided, and so on. Between the Mission and Summary I hope that a reader can see what I initially intended and what actually

69.3%, OK?

The Association for Software Testing is crowd-sourcing a book, Navigating the World as a Context-Driven Tester , which aims to provide responses to common questions and statements about testing from a context-driven perspective . It's being edited by Lee Hawkins who is posing questions on Twitter ,  LinkedIn ,  Slack , and the AST mailing list and then collating the replies, focusing on practice over theory. I've decided to contribute by answering briefly, and without a lot of editing or crafting, by imagining that I'm speaking to someone in software development who's acting in good faith, cares about their work and mine, but doesn't have much visibility of what testing can be. Perhaps you'd like to join me?   --00-- "What percentage of our test cases are automated?" There's a lot wrapped up in that question, particularly when it's a metric for monitoring the state of testing. It's not the first time I've been asked either. In my

Why Do They Test Software?

My friend Rachel Kibler asked me the other day "do you have a blog post about why we test software?" and I was surprised to find that, despite having touched on the topic many times, I haven't. So then I thought I'd write one. And then I thought it might be fun to crowdsource so I asked in the Association for Software Testing member's Slack, on LinkedIn , and on Twitter for reasons, one sentence each. And it was fun!  Here are the varied answers, a couple lightly edited, with thanks to everyone who contributed. Edit: I did a bit of analysis of the responses in Reasons to be Cheerful, Part 2 . --00-- Software is complicated, and the people that use it are even worse. — Andy Hird Because there is what software does, what people say it does, and what other people want it to do, and those are often not the same. — Andy Hird Because someone asked/told us to — Lee Hawkins To learn, and identify risks — Louise Perold sometimes: reducing the risk of harming people —

Testing is Knowledge Work

  The Association for Software Testing is crowd-sourcing a book, Navigating the World as a Context-Driven Tester , which aims to provide responses to common questions and statements about testing from a context-driven perspective . It's being edited by Lee Hawkins who is posing questions on Twitter ,  LinkedIn ,  Slack , and the AST mailing list and then collating the replies, focusing on practice over theory. I've decided to contribute by answering briefly, and without a lot of editing or crafting, by imagining that I'm speaking to someone in software development who's acting in good faith, cares about their work and mine, but doesn't have much visibility of what testing can be. Perhaps you'd like to join me?   --00-- "We need some productivity metrics from testers" OK. I'd like to help you meet your need if I can but to do that I'll need to ask a few questions. Let's start with these: Who needs the metrics? Is there a particular pr

My Favourite Tool

Last week I did a presentation to a software testing course at EC Utbildning in Sweden titled Exploring with Automation where I demoed ways in which I use software tools to help me to test. Following up later, one of the students asked whether I had a favourite tool. A favourite tool? Wow, so simple but sooo deep!  Asking for a favourite tool could make a great interview question, to understand the breadth and depth of a candidate's knowledge about tools, how they think about an apparently basic request with deep complexity beneath (favourite for what task, on what basis, in what contexts, over what timescale?  what is a tool anyway?) and how they formulate a response to take all of that into account. I could truthfully but unhelpfully answer this question with a curt Yes or No. Or I could try and give something more nuanced. I went for the latter. At an extremely meta level I would echo Jerry Weinberg in Perfect Software : The number one te

Trying to be CEWT

I attend, enjoy, hopefully contribute to, and get a lot from, the local tester meetups and Lean Coffee  in Cambridge. But I'd had the thought kicking around for a long time that I'd like to try a peer workshop inspired by MEWT , DEWT , LEWT and the like. I finally asked a few others, including the local meetup organisers, and got mostly positive noises, so I decided to give it a go. I wrote a short statement to frame the idea, based on LEWT's: CEWT ( Cambirdge Exploratory Workshop on Testing ) is an exploratory peer workshop. We take the view that discussions are more interesting than lectures. We enjoy diverse ideas, and limit some activities in order to work with more ideas. and proposed a mission for an initial attempt to validate it locally on a small scale. Other local testers helped to refine the details in usual the testing ways - you know: criticism, questions, thought experiments, challenges, comparisons, mockery and the rest - and a list of potential at

Fail Here or Fail There

The First Law of Systems-Survival, according to John Gall, is this: A SYSTEM THAT IGNORES FEEDBACK HAS ALREADY BEGUN THE PROCESS OF TERMINAL INSTABILITY Laws are all-caps in Systemantics . Not just laws, but also theorems, axioms, and corollaries. There are many of them so here's another (location 2393-2394): JUST CALLING IT “FEEDBACK” DOESN’T MEAN THAT IT HAS ACTUALLY FED BACK There was a point when I realised, as the capitalised aphorisms rolled by, that I was sinking into the warm and sweetly-scented comforting foamy bathwater of confirmatory bias. Seen, seen, seen! Tick, tick, tick! I took the opportunity to let myself know that I'd been caught in the act, and that I needed to get out of the tub and start to challenge the content.  Intervening at that moment was congruent: I was in a context where I would accept it and prepared to change because of it. Of course, I enjoyed the deep irony of nodding along with Gall when he talked about

Testing and Words

  The other day I got tagged on a Twitter thread started by Wicked Witch of the Test about people with a background in linguistics who’ve ended up in testing. That prompted me to think about the language concepts I've found valuable in my day job, then I started listing them, and then realised how many of them I've mentioned here over the years .   This post is one of an occasional series collecting some of those thoughts.  --00-- In The Complete Plain Words , Ernest Gowers notes, acidly, that: What appears to be a sloppy or meaningless use of words may well be a completely correct use of words to express sloppy or meaningless ideas. It surely sounds trite to say it but our choice of words can make a significant difference to how well our message is understood, and how we are judged. We choose from amongst those words we know, our lexicons . The more my lexicon agrees with yours, the greater our chance of us achieving a shared understanding when we converse. But lexic

Use the Force Multiplier

On Fridays I pair with doctors from Ada 's medical quality team. It's a fun and productive collaboration where I gain deeper insight into the way that diagnostic information is encoded in our product and they get to see a testing perspective unhindered by domain knowledge. We meet at the same time each week and decide late on our focus, choosing something that one of us is working on that's in a state where it can be shared. This week we picked up a task that I'd been hoping to get to for a while: exploring an API which takes a list of symptoms and returns a list of potential medical conditions that are consistent with those symptoms.  I was interested to know whether I could find small input differences that led to large output differences. Without domain knowledge, though, I wasn't really sure what "small" and "large" might mean. I prepared an input payload and wrote a simple shell script which did the following: make a