When the Smart Software Makes You Feel Very Dumb

The so-called smart tools I have access to should have shaved down a thirty-minute task to six minutes. Instead, I wasted three hours. How smart is that -- and what does my experience say about how software helps or hinders work?

On Friday, I decided to try and take advantage of a feature Hootsuite offers to its paying customers: the ability to bulk-upload individual tweets to a Twitter account. The benefit to this would be, presumably, time-saving: Instead of individually scheduling each tweet (a 30-second process) and spending maybe 30 minutes doing a weekend's worth of Twitter uploads, I could spend five minutes doing a spreadsheet and converting it to a .CSV file, then upload it to HootSuite over another 60 seconds, and et voila! Smart tools would help save me time.

I had not counted on the productivity-obscuring features of "smart" and helpful tools at every step.

The first thing I did was read the directions HootSuite had for how to bulk-upload tweets. In them, the site service specifies the format of the spreadsheet you'll convert to a CSV file. They're very clear on how dates and times must be formatted:

Go on, guess how often I had to hit command-1, click "Numbers," then adjust the date and time formats in Microsoft Excel. This is because Microsoft Excel 2010 for the Mac will helpfully format dates or times in a way it thinks works better, no matter how often you specify otherwise.

Eventually, I decided to see if Google Sheets would listen to my explicit directions -- always include a two-digit value for dates, months and hours, and use a 24-hour clock -- and discovered that if I was able to format the date, I lost the time and vice-versa.

Remember, these are spreadsheets that are helping you by automagically formatting your data for you.

So I figured I'd just export the spreadsheet as a .CSV file and do a simple search-and-replace for the offending date or time formats. It was more effective for me to craft a search query in my text editor and use a simple tool than it was for me to use the preferences and features of software that was supposed to help me work smarter.

I double-checked my .CSV file against the Hootsuite instructions, attempted to upload it ... then learned that my tweets were riddled with errors, every single one. However, I did not learn what those errors were or how I could fix them. The error message only identified the errors as being in the body of the tweet.

It took me ten minutes of experimentation to discover that Hootsuite's spreadsheet-uploading tool will not accept any tweet that has a colon in this.

(And why did all my tweets have colons in them? Because another "smart" tool in Twitter -- and, by extension, Hootsuite -- prevents someone from scheduling or posting duplicate tweets. While I applaud the general sentiment behind this, on a practical level, it means that if I want to give readers multiple opportunities to see one of Rich Hay's great stories, I have to find multiple ways to write the same tweet for different times of the day. As of right now, we introduce the tweets with something like "Your morning read: Become an Xbox insider" or "ICYMI: Become an Xbox insider.")

So. I had to find a way to work around Twitter and Hootsuite's smart detection of duplicate tweets and another way to work around Hootsuite's undocumented refusal to accept tweets with colons in them. I rewrote all the tweets, tried to upload the .CSV ...

And remembered right then that editing the tweets in the spreadsheet meant I had to re-format all the dates and times my spreadsheet kept helpfully reformatting against my preferences.

Once I had checked my .CSV file against the Hootsuite instructions to ensure the dates and times were correct and to make sure I had no colons in any of my tweets, I uploaded the file, confident in its success ...

... And got an error message saying that Hootsuite's URL shortening tool was having difficulty with one of my tweet URLs -- no word on which one -- and I might want to check them all before uploading.

I checked all three dozen URLs. They all worked. I went and shortened them, just in case the lengthy URLs associated with our galleries was throwing off the syntax somehow. I changed the spreadsheet to include the shortened URLs, exported it as a .CSV file, reformatted the dates and times, then tried to upload the .CSV file again.

Still no success. 

At this point, I had spent nearly three hours carefully and slowly checking three dozen tweets to see why I couldn't schedule them en masse. At every step of the production process, from my spreadsheet apps (I was doing this in both Google Sheets and Microsoft Excel to see if one spreadsheet tool was easier to format than the other), to the URL shortener, to Hootsuite, some "smart" feature meant to make my work easier had created more work. I should have just sunk a half hour into writing and scheduling tweets in Hootsuite -- tweets, by the way, I can compose with both colons and original URLs -- rather than use the time-saving smart tool.

Part of the problem rests with Hootsuite; every time the service rejected my .CSV file, it gave an error message, but no pointer on why the error was triggered or how to fix it. Error messages should not be the pop-up window equivalent of street signs; they should be the pop-up window equivalent of directions. The former tells you where you are at the moment, but the latter tells you how to get to where you want to be.

But a larger portion of the problem rests with the premise that we can build "smart" behavior into tools to save us time. The problem with this premise is that a lot of the "smart" stuff -- autocorrect or autoformat or detection of specific patterns -- is encoded into the tool and is not very flexible or editable. 

In a way, the failure of these tools reminds me of Alton Brown's old television show, Good Eats. In it, Brown advised against kitchen unitaskers, i.e. those gadgets and gizmos that do only one thing, and in only one way. It was better for users to develop skills that would let them use a suite of simple, flexible tools rather than remain comparatively unskilled and deal with a plethora of kitchen clutter.

I thought about the "no unitaskers!" rule when I thought over the afternoon of wasted work. The one tool that worked consistently for me and did what I needed it to, every time I needed it to, was a simple search function. As a tool, it's incredibly basic: All it does is look for exactly what the user specifies. The power rests in the user's ability to craft the queries they need. And a user develops those abilities by learning a few basic rules, then using their own smarts to figure out when and how to apply those rules to a particular task.

Unless and until "smart" tools in our lives reject the unitasker model and move toward flexible and conditional use, they're not going to be smart and they're not going to help us learn how to improve the quality of our own work.

But I have hope that some tools moving in this direction. For example, the autocorrect feature on my iPhone has demonstrated that it's capable of learning and filling in my frequently-typed phrases. That's smart -- it saves me time and it gives me the option of fixing my errors or teaching it that no, I meant to type "Subaru" instead of "suburban." And from what I've seen in demos on machine learning, the very point to machine learning is to create a set of software functions that are flexible, adaptive and individualized.

So long as we stop equating "smart" with "automatic" and start associating it with "modifiable" and "situational," we'll be good. Let's get smart about smart tools.