Posts in the Culture Category

The Broken Physics of “The Umbrella Academy” Finale

Published 4 years, 10 months past

Not long ago, Kat and I got around to watching The Umbrella Academy’s first season on Netflix.  I thought it was pretty good!  It was a decent mix of good decisions and bad decisions by people in the story, I liked most of the characters and their portrayals, and I thought the narrative arcs came to good places. Not perfect, but good.

Except.  I have to talk about the finale, people.  I have to get into why the ending, the very last few minutes of season one, just didn’t work for me.  And in order to do that, I’m going to deploy, for the first time ever, a WordPress Spoiler Cut™ on this here blog o’ mine, because this post is spoilerrific.  Ready?  Here we go.

Massive, massive spoilers and a fair amount of science ahead!

Results of the GIF Survey

Published 7 years, 4 months past

The GIF Survey is complete.  In just under a week, 1,457 people gave their answers on how they pronounce the acronym, and their perceptions of the rightness of that pronunciation.  I thought that, today of all days, it made some sense to share the results of a far less momentous poll.

For those who missed, it, how this survey worked was that the first question was: “How do you pronounce GIF?”  To this, the choices were:

  • The obviously correct way
  • The clearly incorrect way

Upon answering this, respondents moved on to a section that asked three optional demographic questions: age, gender, and race/ethnicity, all as open text fields.  These had about a 16% skip rate, and about a 4% ‘faithless’ response rate; that is, answers that were clearly jokes, insults, or other explicit refusals to answer the question as intended.

Once the demographic questions were answered or skipped, there was a final question: “How do you pronounce GIF?”, exactly the same as the first question of the survey.  Only this time, the options were:

  • Hard G (like “gift”)
  • Soft G (like “gin”)

For both pronunciation questions, the answer order was randomized so as to avoid any first-choice advantage.  The demographic questions, being open entries, didn’t have options to randomize.

(Aside: I discovered in the course of the survey that there are other pronunciations, most commonly in non-English languages.  My apologies to those who fell outside the binary choice I presented.)

So!  The results came out like this:

Table 1. Perception of pronunciation
The obviously correct way 83.7%
The clearly incorrect way 16.3%

First of all, it amuses and slightly mystifies me that more than 16% of respondents feel they say it the “incorrect” way.  Second of all, these percentages didn’t line up with actual pronunciation.

Table 2. Actual pronunciation
Hard G 77.8%
Soft G 22.2%

This deserves a closer look.  How do perceptions of correctness break down by actual pronunciation?

Table 3. Perception versus pronunciation
Pronunciation “Correct” “Incorrect”
Hard G 87.3% 12.7%
Soft G 71.2% 28.8%

In other words, people who pronounce it with a hard G are significantly more likely to believe their pronunciation is correct than those who go the soft-G route.

It’s an interesting inversion of what one might (perhaps naïvely?) expect: given that the creator of the format has explicitly said the pronunciation is with a soft G, one might expect that those who use the hard G know it’s incorrect but say it anyway.  My personal opinion is that this is actually a reflection of human nature: faced with evidence that undermines our instinctive reactions, we tend to double down.  (Of course, if the evidence lines up with what we believe, we seize on that too.)

Now: demographics, which actually were the point of the survey, but not in the way I think some people assumed.  After I did my first, tongue-in-cheek version of the poll on Twitter, my colleague Aki noted that she’d love to know something about the demographics behind those results, something I’d had flitting around in the back of my mind.  Her comment made me decide to just go for it.  What I wanted to see was whether there were significant differences in perceptions of correctness in various groups.  For example, one might hypothesize that those identifying as female were more likely to say their choice was incorrect.  Well, if that were the hypothesis, what evidence I was able to gather contradicts it.

Table 4. Perception of pronunciation by gender
Gender “Correct” “Incorrect”
Female 83.4% 16.6%
Male 83.5% 16.5%

Roughly speaking, of those people who gave an answer about their gender (81.5% of the total), about 25% of respondents identified as female, and about 70% identified as male.  One thing that did jump out at me was that those identifying as female were more likely to use the hard G, rather than the soft G.  Not by a lot, possibly within the margin of error, but still.

Table 5. Actual pronunciation by gender
Gender Hard G Soft G
Female 82.7% 17.3%
Male 77.2% 22.8%

The other thing that interested me was how patterns of pronunciation and correctness would correspond, if they did at all, to age — for example, were younger respondents more or less likely to think they were right than older respondents?  I decided to group by decades, in effect.  Of the 81.6% of respondents who gave a reasonably valid age (I tossed, for example, “1.7977E+308”), here’s how they clustered.

Table 6. Age groups
20-29 22.2%
30-39 42.7%
40-49 25.5%
50-59 6.6%

There weren’t enough respondents outside the 20-59 range to analyze.  I’m not even sure about the 50-59 group, to be honest — I’m not sure 79 replies out of 1,457 is enough.  But what the heck, I’m rolling with it.  Respondents’ perception of correctness didn’t change a lot, but did seem to rise a bit with age.

Table 7. Perception by age group
Age Group “Correct” “Incorrect”
20-29 81.8% 18.2%
30-39 84.3% 15.7%
40-49 83.2% 16.8%
50-59 86.1% 13.9%

It would be interesting to see if a different division of age groups would create different results.  But what really caught my eye was how the pronunciation shifts with age: younger respondents were notably more likely to use the soft G than older respondents.

Table 8. Pronunciation by age group
Age Group Hard G Soft G
20-29 73.1% 26.9%
30-39 77.8% 22.2%
40-49 84.2% 15.8%
50-59 83.7% 16.5%

So if you’re a soft-G speaker and are convinced that’s correct, perhaps you can take comfort in the belief that the children are our future.

I’m not going to present numbers on race/ethnicity.  This is partly because the question was a MacGuffin: I asked it because it would have seemed odd not to after asking for age and gender, and also because I’ve found over the years that asking for ethnic or racial identification is a handy way to give some people a chance to vent a little built-up animus.  The other reason is that even after filtering out the few abusive and the somewhat more numerous “decline to answer” replies, the remaining values are all over the place and difficult to make consistent.

If you’d like to try, you can download the filtered-and-resorted data set for that question, as well as similar sets for age and for gender.  I’ve also put up a data set containing just the answers to the two mandatory pronunciation questions.  Feel free to analyze them as you like!  Each file is a ZIP of the data set in tab-separated text format, so they’re pretty small.

And just to be clear, I’m not planning to post the complete data set, just in case any combination of demographic answers could be used to reconstruct an identity.  (Each set was sorted differently, so a line number in one set doesn’t correspond to the line number in another.)

So what did all this tell us?  It told us something about the people who saw the survey and chose to respond.  It told us that if the results are representative, then people who are older tend to use the hard G and be more convinced of their rightness.  Maybe that’s representative of the world as a whole, and maybe not.  It may not mean a lot in the grand scheme of things, but it was fun to ask, hopefully fun for people to answer, and fun to crunch the numbers that resulted.

My thanks to everyone who took part, and to Aki for prompting me to do it in the first place.


A More Compassionate Facebook

Published 8 years, 5 months past

It’s been a busy couple of weeks for Facebook, in terms of compassionate design decisions.

First they announced that they aren’t adding a Dislike button, but they are adding a set of six emoji reactions to the “Like” button, so you can indicate a wider range of emotion.  Some people immediately linked this to Slack, as if emoji reactions hadn’t been a thing on social media for the last couple of years.  I happened to see Sally Herships asking “what are your thoughts?” about it on Twitter (heh), and oh, I had thoughts.  I ended up sharing some of those thoughts by phone, and one of them was part of a segment on American Public Media’s Marketplace.

It’s funny, in a way, that my thought on marketing and advertisers was what made it into the piece, because I think that was literally my whole thought about that side of things.  Most of the rest of my conversation with Sally was about how Facebook could use these reactions as a way to avoid insensitive design choices.  As an example, a status update that gets lots of interaction in the frowny-face or sad-face realm could be avoided when it comes to things like Year in Review.  I said something to the effect of:

People are sharing everything about their lives, positive and negative, billions of us every day.  That isn’t going to stop, so it’s great to see Facebook making changes to meet us where we are, or at least meet us partway.

These reaction emoji almost certainly aren’t the last word on this, but they’re a credible initial attempt.  In more than one sense, they’re a first step into a larger world.


Next, Facebook introduced filtering for its On This Day (OTD) feature.  This is another step in the evolution of On This Day, one that’s very welcome.  Facebook had already been revising its language to be more humane, shifting from simple “Relive this memory” to nuanced language expressing care and openness.

The original and more recent copy at the top of an On This Day memory.

With its new OTD preferences, Facebook now lets you define ranges of dates you’d like to be blacklisted, in effect, as well as people you don’t want to see memories about.  I’d commented on the lack of this, back when OTD launched:

…what I notice here is what’s missing:  I don’t see any reference to an ability to opt out of On This Day, either for certain days or altogether.

So far as I can tell, you still can’t opt out entirely; even if you turn off all notifications, you can still get memories inserted into your timeline.  For me, I see about one a month, more or less.  But here’s the interesting thing: they’re almost never my memories.  In what I still regard as a major gamble by Facebook, On This Day will show you posts, pictures, and videos posted by someone else, but on which you were tagged.  I presume (though I have no simple way to test) that adding a person in the OTD filtering preferences will prevent you from seeing memories in which they’re tagged as well as memories they posted.

If so, that’s a really smart step, as I can only imagine how a spiteful ex might abuse OTD.  It still leaves open the possibility of old posts that you don’t remember being tagged on suddenly appearing.  In many cases, that will be a delightful moment, but in many others, the exact opposite of that.  This is why I regard Facebook’s decision to show you posts from other people a gamble.  Even if they show unwanted memories to just 1% of their user base — a ridiculously low percentage — that’s literally 10 million people a day.

Still: wrinkles or no, flaws or no, the presence of filtering preferences is a major enhancement to On This Day.  I could block out all of June 2014, if I so chose.  There might be years where I blocked it, and others where I removed the block.  The important thing is that I’m being given that capability, in an environment that’s already designed to show me memories and acknowledge that it’s easy to get that wrong.  The user experience for adding filters is still clunky, but much like the reaction emoji, I view this as a credible first try, not the final word.

All this has made for some interesting Slack discussions between me and Sara, as we literally just finished the manuscript for our forthcoming, still-not-quite-titled-but-we’re-really-close-honest book on compassion in design.  Which has references to things like On This Day, so we’re already revising a book that hasn’t even been published yet.  And when will it be published?  We’re pulling for early next year, which sounds like a long way away until you remember that 2015 is getting close to done.

Kudos to Facebook, both for its efforts to be kinder in what they do and for its willingness to try.  Not many businesses, let alone social-media titans, have had the courage to think about what can go wrong in this realm, let alone actually acknowledge missteps and work to do better.  Well done.


The Shape of Things to Come

Published 8 years, 6 months past

Software may be eating the world, but we are shaping it.  What we do now — what we build, how we act, what we tolerate — will profoundly influence how society develops over the next few generations.

That’s not because what happens now will change you or me.  We’re unlikely to change much, if at all.  We’re set in our ways, most of us.

Our children are not.

What they see online will seem normal to them, just as what we saw growing up seemed normal to us.  And because there is no meaningful distinction between online and offline, what they come to accept as normal online will be seen as normal offline.

So the way we build our networks matters in the most profound possible way.  If we build networks that make it easy to abuse and harass, and make it difficult to defend against abuse and harassment, our children will come to see that as normal, even desirable.  Similarly, if we build networks where it’s hard to abuse and harass, and easy to defend against such attempts, that will become the norm.

System design is social design.  The question is, what kind of society do we want to design?

And the more important question is, when are we going to start?

This article was originally published at The Pastry Box Project on 2 September 2015.


Dislike

Published 8 years, 6 months past

Facebook is emotionally smarter than we give it credit for, though perhaps not as algorithmically smart as it could be.

I’ve been pondering this for a few weeks now, and Zeynep Tufekci’s “Facebook and the Tyranny of the ‘Like’ in a Difficult World” prodded me to consolidate my thoughts.

(Note: This is not about what Tufekci writes about, exactly, and is not meant as a rebuttal to her argument.  I agree with her that post-ranking algorithms need to be smarter.  I also believe there are design solutions needed to compensate for the unthinking nature of those algorithms, but that’s a topic for another time.)

Tufekci’s piece perfectly describes the asymmetrical nature of Facebook’s “engagement” mechanisms, commented on for years: there is no negative mirror for the “Like” button.  As she says:

Of course he cannot like it. Nobody can. How could anyone like such an awful video?

What happens then to the video? Not much. It will mostly get ignored, because my social network has no way to signal to the algorithm that this is something they care about.

What I’ve been thinking of late is that the people in her network can comment as a way to signal their interest, caring, engagement, whatever you want to call it.  When “Like” doesn’t fit, comments are all that’s left, and I think that’s appropriate.

In a situation like Tufekci describes, or any post that deals with the difficult side of life, comments are exactly what’s called for.  Imagine if there were a “Dislike” button.  How many would just click it without commenting?  Before you answer that question, consider: how many click “Like” without commenting?  How many more would use “Dislike” as a way to avoid dealing with the situation at hand?

When someone posts something difficult — about themselves, or someone they care about, or the state of the world — they are most likely seeking the support of their community.  They’re asking to be heard.  Comments fill that need.  In an era of Likes and Faves and Stars and Hearts, a comment (usually) shows at least some measure of thought and consideration.  It shows that the poster has been heard.

Many of those posts can be hard to respond to.  I know, because many of the Facebook posts my wife and I were making two years (and one year) ago right now were doubtless incredibly hard to read.  I remember many people leaving comments along the lines of, “I don’t know what to say, but I’m thinking of you all.”  And even that probably felt awkward and insufficient to those who left such comments.  Crisis and grief and fear in others can make us very uncomfortable.  Pushing past that discomfort to say a few words is a huge show of support.  It matters.

Adding “Dislike” would be a step backward, in terms of emotional intelligence.  It could too easily rob people who post about the difficult parts of life of something they clearly seek.


Warning Hashflags

Published 8 years, 10 months past

Over the weekend, I published “Time and Emotion” on The Pastry Box, in which I pondered the way we’re creating the data that the data-miners of the future will use to (literally) thoughtlessly construct emotional minefields — if we don’t work to turn away from that outcome.

The way I introduced the topic was by noting the calendar coincidence of the Star Wars-themed tradition of “May the Fourth be with you” and the anniversary of the Kent State shootings in 1970, and how I observe the latter while most of the internet celebrates the former: by tweeting some song lyrics with a relevant hashtag, #maythe4th.  I did as I said I would…and Twitter blindly added a layer of commentary with a very simple little content filter.  On twitter.com and in the official Twitter app, a little Stormtrooper helmet was inserted after the hashtag #maythe4th.

So let’s review: I tweeted in remembrance of a group of National Guardsmen firing into a crowd of college students, wounding nine and killing four.  After the date hashtag, there appeared a Stormtrooper icon.  To someone who came into it cold, that could easily read as a particularly tasteless joke-slash-attack, equating the Guardsmen with a Nazi paramilitary group by way of Star Wars reference.  While some might agree with that characterization, it was not my intent.  The meaning of what I wrote was altered by an unthinking algorithm.  It imposed on me a rhetorical position that I do not hold.

In a like vein, Thijs Reijgersberg pointed out that May 4th is Remembrance of the Dead Day in the Netherlands, an occasion to honor those who died in conflict since the outbreak of World War II.  He did so on Twitter, using the same hashtag I had, and again got a Stormtrooper helmet inserted into his tweet.  A Stormtrooper as part of a tweet about the Dutch remembrance of their war dead from World War II on.  That’s…troublesome.

Michael Wiik, following on our observations, took it all one step further by tweeting a number of historical events collected from Wikipedia.  I know several of my British chums would heartily agree with the 1979 tweet’s added layer of commentary, but there are others who might well feel enraged and disgusted.  That could include someone who tweets about the election in celebration, the way people sometimes do about their heroes.

But what about appending a Stormtrooper helment to an observance of the liberation of the Neuengamme concentration camp in 1945?  For that matter, suppose someone tweets May-4th birthday congratulations to a Holocaust survivor, or the child of a Holocaust survivor?  The descendant of a Holocaust victim?

You might think that this is all a bit much, because all you have to do is avoid using the hashtag, or Twitter altogether.  Those are solutions, but they’re not very useful solutions.  They require humans to alter their behavior to accommodate code, rather than expecting code to accommodate humans; and furthermore, they require that humans have foreknowledge.  I didn’t know the hashtag would get an emoji before I did it.  And, because it only shows up in some methods of accessing Twitter, there’s every chance I wouldn’t have known it was there, had I not used twitter.com to post.  Can you imagine if someone sent a tweet out, found themselves attacked for tweeting in poor taste, and couldn’t even see what was upsetting people?

And, as it happens, even #may4th wasn’t safe from being hashflagged, as Twitter calls it, though that was different: it got a yellow droid’s top dome (I assume BB-8) rather than a Stormtrooper helmet.  The droid doesn’t have nearly the same historical baggage (yet), but it still risks making a user look like they’re being mocking or silly in a situation where the opposite was intended.  If they tagged a remembrance of the 2007 destruction of Greensburg, Kansas with #may4th, for example.

For me, it was a deeply surreal way to make the one of the points I’d been talking about in my Pastry Box article.  We’re designing processes that alter people’s intended meaning by altering content and thus adding unwanted context, code that throws pieces of data together without awareness of meaning and intent, code that will synthesize emotional environments effectively at random.  Emergent patterns are happening entirely outside our control, and we’re not even thinking about the ways we thoughtlessly cede that control.  We’re like toddlers throwing tinted drinking glasses on the floor to see the pretty sparkles, not thinking about how the resulting beauty might slice someone’s foot open.

We don’t need to stop writing code.  We do need to start thinking.


Time and Emotion

Published 8 years, 10 months past

This coming Monday, as has become tradition, a significant fraction of the Twitter user base will send out Star-Wars-themed tweets tagged #maythe4th or #maythefourthbewithyou, because saying the day in that way makes for a handy bit of wordplay.  There will be cosplay pictures, Yoda-esque inversions of sentence structure, and probably (this year) a fair bit of squeeing about the upcoming sequel and its brilliantly fan-service trailer.

Also this coming Monday, as has become tradition for me, I will send out a tweet containing the opening lines of “Ohio” by Crosby, Stills, Nash & Young, tagged #maythe4th, because it was on May 4th, 1970 that National Guardsmen fired a volley into a crowd of Kent State college students, wounding nine and killing four.

Anniversaries are potent psychological markers.  We reflect on historical events, both global and personal, that have particular meaning to us.  We celebrate the days of our birth, of first meeting our loved ones, of all manner of wonderful life-changing moments.  We mourn the days of our losses, of our betrayals, of all manner of terrible life-changing moments.  In every heart, a secret calendar.

There are only so many days in the year; pile enough things together on a calendar, and some of them will coincide.  Some of those alignments will coalesce into rays of remembered joy, warming us from the past.  Others will form spears of relived pain, lodging afresh in our hearts.  A few may do both, comforting and piercing all at once.

The longer we spend online, the more traces of those secret calendars will take public shape.  The dates of my first marriage and divorce are not, to the best of my recollection, recorded anywhere online, but the date of my second (and current) marriage is there, thanks to some early blog posts.  The date of my first professional award is there.  The dates of our children’s placements and adoptions are there.  The dates of my daughter’s illness and death are there.

The more we build online networks, not physical networks but social and emotional networks, the more pieces we leave lying around for algorithms to gather together and present to us with no real thought for what those pieces actually mean, or for how they should or shouldn’t fit together.  A human can glance through a pile of photos and tell which are emotionally or even narratively out of place.  Code cannot.  A human can quickly determine which scraps of text and pixels were happy at the moment of their creation, only to be transformed into talismans of sorrow by later events.  Code cannot.

We’re collectively creating strata of data, adorned with easy bits of metadata like time and date and sometimes place, but lacking all the truly important metadata like feeling and meaning.  As we share with each other, we share with the future.  We share with the companies that help us share with each other, because it’s easy to store it all.  Content in the old network was ephemeral, and in the older networks was tangible but private.  In the new networks, everything we create is easy to retrieve — if not for us, as users of the network, then at least for the code that runs on the same machines which accept all that we share.

And so, more and more with every passing day, code is written to reach back into everything we’ve created, assembling it along easily-identified axes like Likes or Faves or geographic coordinates or the day of the year, in order to show it to us again.  Sometimes it’s code we invite into our lives, but not always.  Sometimes we find the code that drives the networks we use resurrecting our past without warning.

This will not always be welcome.

There are things we can do to make our remorselessly remembering routines more humane, and most of those things are rooted in experience design.  We can design compassionate consent requests ahead of introducing new functionality, and easy ways to mark which dates and memories and bits of data should be avoided, and even design thoughtful expressions of remorse and apology.  We can and should add this very human layer of thoughtfulness to cushion us from literally unthinking code that yields results which may harm as easily as they may heal.

It won’t be easy, and we’ll make mistakes no matter how hard we try.  Our very attempts to be thoughtful may backfire and make things worse, but we’ll learn from those mistakes and do better the next time.

Nothing could be more human than that.

This article was originally published at The Pastry Box Project on 2 May 2015.


Taking Steps

Published 9 years, 2 weeks past

Not too long ago, I got Step Nined on Facebook.

If that didn’t parse as regular English for you, Step Nine is part of the twelve-step program offered by Alcoholics Anonymous. It states, with edits for out-of-context clarity:

“[Make] direct amends to [people you have harmed] wherever possible, except when to do so would injure them or others.”

And so someone I knew back in the town where I grew up, a classmate all throughout my pre-college education who I’ll call John, sent me a private message on Facebook apologizing for how he’d treated me, stating that he’d had no reason other than just having been a mean kid, and hoping that things were going well for me and my family.

I’ve pondered this message quite a bit in the interim. The message brought me neither upset nor relief, though I can well imagine that many people in such a situation would feel one or both. I bore no mental or physical scars with his signature upon them. There was no need of closure, or of re-opening, or really of anything, at least from my point of view. He and his actions toward me, positive or negative, are one thin thread in the complex skein that was my childhood, lost in the overall pattern.

In truth, John’s message aroused more pity in me than anything else. I thought as I read it, What must he have endured as a child, that hurting other people seemed normal to him?  And in that thought, I felt an echo from the past, as though the question had come to me before. Perhaps my parents made the observation, as I struggled through growing up, and I was finally able to hear it now. I’m not sure. It doesn’t really matter. If his message is anything to go by, whatever John did has been far more damaging to him than it ever might have been to me.

Still, I keep coming back to John’s message and pondering it further. What I’ve thought about, far more than its contents or the history it references, has been the simple fact of how it happened, and what that means.

Had John wanted to offer amends in, say, the late 1990s, he would have had to actively seek me out. It would have taken the effort of calling my parents to ask for contact information, or other people he thought might have it, and then making that call to me. The social distance would have been a barrier to contact, one whose surmounting signified the importance of the act to him. And then, when he did make that call, he would have talked to me, able to gauge my reaction. There would have been a feedback loop to tell him whether or not his amends were injurious in some way.

And yes, of course, John could have done exactly that today. He could have kept his process entirely off Facebook and gone through those efforts, as an act of personal penance or just as a useful social signifier. Or, perhaps, he could have contacted me on Facebook to ask for my phone number, with a brief statement as to why he was asking for it, and then let my decision to allow the contact or not be a measure of whether it would in fact be injurious.

But he didn’t. Because the internet has disintermediated social effort.

What I wonder about, as I ponder this small signal, is the depth of his remorse. How much does John really mean it, and how much is he going through the motions, trying to get through Step Nine as quickly as possible so that he can reach Step Twelve sooner? Is he working through his personal pain, or is he grinding the leveling process? It’s impossible for me to say. I know it’s a lot easier to send a bunch of “sorry” messages to your contacts than it is to talk to each person you feel you’ve wronged, one on one, one by one, and go through that painful process over and over and over again.

I wonder if that simple ease of contact has robbed him of a critical component of his healing process.

Or, if you want to be more accurate, I wonder if that simple ease of contact lured him into a course of action that was harmful to his healing process.

The things we build are almost always meant to make things faster, more efficient, easier. Perhaps, sometimes, they should be harder.

This genie will not go back into the bottle. The internet isn’t going away and Facebook still has a long way to go before its fall. Even then, something will have replaced it. There’s no reason to think these sorts of connections will become more difficult to make, technologically speaking.

I wonder if they will become more difficult, socially speaking — if an act like that will become frowned upon, as we might frown upon a form letter condolence note. I wonder what sort of protocols and expectations, what social mores, will emerge over time in response to the disruptions our work has caused and will cause, and how they will shape personal interactions at all levels.

I wonder how much effort we should be putting into influencing the evolution of those emergent social constructs, whether through our work or our personal interactions, and how much of that effort would be ultimately fruitless.

I wonder how intentional people are about what they do, online as well as off; and how intentional they should be.

I wonder what I should say to John.

This article was originally published at The Pastry Box Project on 2 March 2015.


Browse the Archive

Earlier Entries