If you’ve spent any time reading on the web the past week, odds are you’ve read something written by a robot—and you didn’t even realize it.
Robot writers are algorithms that collect and analyze data and then turn them into readable narratives. Many news sites like the Los Angeles Times and Forbes are already using them. Even Wikipedia has articles that weren’t written by humans.
Reception to robot journalists has been mixed: Some see how robots can be useful, while others take this as another sign that creative professionals are being devalued.
To see for ourselves whether these robot writers can be useful or how they might impact creative work, we can start by getting to know one of them.
Meet the robot journalist that writes for AP and Yahoo
Created by Automated Insights, Wordsmith is a platform that can take any kind of structured data and generate a plain English narrative from it. Since it doesn’t need breaks, days off, or long research hours, it can produce 2,000 simple articles per second. It is on track to producing over 1 billion stories by the end of the year.
What might be surprising to most is that Wordsmith’s “voice” isn’t completely robotic—it can adapt, depending on the context. For example, it’s written some objective, matter-of-fact pieces, such as this Radio Shack corporate earnings report for the Associated Press. But it can also generate more lively, conversational pieces, such as the personalized, trash-talking Fantasy Football recaps for Yahoo.
However, there are significant limits to its powers.
Can a robot be a storyteller and thought leader?
Though robot journalists can construct certain types of narratives from data, it might have trouble with a key component of publishing: storytelling and thought leadership.
“I am the person at the company who probably does most of our thought leadership pieces,” said James Kotecki, manager of media and public relations for Automated Insights. “It’s certainly not the case that a robot or algorithm has taken my job.”
As an example, Kotecki cites the automated corporate earnings reports generated by Wordsmith for the Associated Press. These short reports provide details about a company’s gains and losses, how they fared compared to analyst predictions, and how the stock price is doing. As a result, their journalists no longer have to deal with mindless number crunching, data processing, and formulaic writing.
“Then if you want to have any additional context or qualitative analysis, you need to have a human as part of that,” Kotecki said. “That means that the journalists involved are still going to go out there and get a quote from the CEO, they’re still going to put whatever happened in their earnings report into a broader context.”
In that sense, a robot writer can serve as a useful editorial assistant.
Putting data into a broader context is especially important for brands where humanitarian and environmental issues are a key aspect of their messaging. Such brands can range from nonprofits like charity:water to companies like Chipotle, whose branding revolves around ethical food production, as well as the restorative narratives depicted in “This Built America,” which was underwritten by Ford Trucks. These approaches need to be compelling and authentic to impact an audience.
This is a significant limitation of robot writers, according to Dr. Noam Lemelshtrich Latar, founding dean of the Sammy Ofer School of Communications at IDC Herzliya. His research focuses on how artificial intelligence and robotics will impact the future of journalism.
“Yes, we can write algorithms that will investigate issues important to human survival, but only the human can think of what should be investigated,” he said. “They cannot think out of the AI box. The AI algorithms will never be able to catch up from these respects. Definitely not to protect our human rights or catch up with changing human values.”
The rise of hyper-scalable personalized content
Earlier this year, we covered three content formats that are about to take off, and one of these formats is personalized content and experiences. Robot writers are a boon to personalized content, since they can automatically segment the content depending on reader data and preferences. Wordsmith, for example, will have created a quarter-billion personalized Fantasy Football recaps by the end of the year. “Not only are your customers getting really engaging, extremely personalized content about what you’re selling or offering to them, but you’ve also collected their data willingly,” Kotecki said.
But the success of these personalized experiences is dependent on data, according to Dr. Lemelshtrich Latar.
“Assuming the digital profiles are accurate, this could have a great effect on the ROI of these businesses, their promotions strategies and their product development and design,” Lemelshtrich Latar said. “Knowing the preferences of their customers will allow these businesses to run effective targeted media campaigns and most important—offer their customers products that will fit their interests even before these customers realize it that these products appeal to them.”
R-O-I, Robot
For brand publishers, robot writers can serve another function: analyzing and reporting marketing metrics. Wordsmith for Marketing, a smaller-scale version of the platform, integrates with Google Analytics and creates client reports for digital marketing agencies.
“Marketing agencies have teams of people where their sole responsibility might be reporting,” said Kieran Wilson, digital marketing manager at Automated Insights. “So instead of having people spend hours and hours in Google Analytics and analyzing and looking at trends and patterns and correlations and producing that report for their clients, it fully automates the process.”
More time for human journalists to “do human stuff”
“Can Wordsmith write a novel?” most people ask Kotecki. The answer is that it’s theoretically possible—if you can figure out a way to structure the data needed to compose a compelling story.
“The problem is that by the time you actually create that massive structured dataset about all the variables that would go into a fictional novel,” he said, “you would probably end up spending way more time creating the dataset and figuring out how to configure Wordsmith to write about it than you would’ve if you just wrote the novel the traditional way.”
In other words, robot journalists can definitely be useful for analyzing data and generating reports on that data in a natural language—and they can do this at scale. But it has serious limitations, and as a result, storytelling is still a human’s game.
“If we’re automating something, it’s going to be better and faster than what a human can do,” said Kotecki, “but that leaves more room actually for humans to do more exciting human stuff.”
|
Showing posts with label web writing. Show all posts
Showing posts with label web writing. Show all posts
Wednesday, September 24, 2014
Does Your Brand Newsroom Need a Robot Writer?
Simplicity, It’s Complicated
“Make everything as simple as possible, but not simpler.”—Albert Einstein
Your smart phone would be simpler if it didn’t have a camera. But it would also be a lot less useful. A paint can is a sublimely simple way to store paint. But it’s a dreadful way to interact with paint: It’s hard to open, hard to close, and guaranteed to make a mess every time you pour it.
Simplicity is an obvious virtue, but it is not the ultimate virtue. Sure, the fewer clicks, taps, steps, hurdles, fields, distractions, and doodads in a UI, the better. We UX designers have a whole bag of tricks for simplifying things—consolidating here, eliminating there, shrinking this, submerging that—and 19 times out of 20, the simpler approach is the better approach.
But, ahhh, that twentieth time. In our enthusiasm for making elements, widgets, and products simple, we need to avoid overshooting the target, making them too simple to serve their purpose well. For example:
- An icon without any words attached to it is nice, simple, and clean. It avoids translation problems, but simultaneously runs a high risk of being incomprehensible in many languages.
- The ubiquitous USB plug fits into a port only if you have the top on top. But since both plug and port are simple rectangles, most people can’t tell top from bottom. You have to examine both closely and compare them carefully to get the top on top. Most people don’t, so we routinely need two or more tries to get the damn thing plugged in. If the mating parts were just slightly more complex—say, shaped like a D or an obtuse isosceles triangle, or with a conspicuous dot always marking the top of both plug and port—we would struggle with USB plugs much less.
- A mobile app that depends on users somehow knowing specialized gestures to navigate has the luxury of presenting a wonderfully simple interface, but a high percentage of users will give up on it before ever stumbling upon those secret handshakes.
- There was an early mobile app that consisted simply of the complete text of the King James Bible—all 750,000+ words of it—with no pesky navigation or search functionality to clutter up the UI (uhhh, I mean, to enable you to find anything). You were left to just scroll and scroll and scroll and …
- There is a service (which shall remain nameless) that syncs up your files on multiple devices. It's designed to automatically and instantly delete a file everywhere if you delete it anywhere. Nice and simple: everything always in sync. And to keep things really simple, there is no warning, no “Are you sure?” and no undo. But 20% of its users view the service as a means of backing up their files. They are in for a nasty surprise when they discover later that their “backup” copy in the cloud has vanished just as soon as any instance of it anywhere else gets deleted, even if that deletion was an accident.
- Many country selector dropdowns consist of a simple list of the world’s 190+ countries in alphabetical order. But such a simple widget will inconvenience every user except those in countries at the top of the alpha list. And chances are most of your users are not in Afghanistan, Albania, or Algeria.
So, yeah, making something too simple can be as problematic as making it too complex, just as making a room too cold can be as uncomfortable as making it too hot.
Scratching the Itch
Simplicity is a one-to-one match between an itch and a scratch. Fine-turning the scratch to the itch usually does mean subtracting, but not always. Sometimes it means a small addition, such as asking users whether they want that file deleted everywhere. To hit the appropriate level of simplicity, we need clarity about exactly what itch we’re trying to scratch.
The goal is not simplicity for its own sake. Simplicity is only a means to an end. The goal is ease. We can admire simplicity as a characteristic, but ease is what we want users to experience.
The goal is not simplicity for its own sake. Simplicity is only a means to an end. The goal is ease.tweet this
“Keep it simple” is a great rule of thumb but a poor reflex ideology, because of all the times it shames us into putting simplicity ahead of ease. Naturally, we want both, and they do tend to go together. But we sometimes have to lessen one to achieve the other. When we do, ease should win. As long as the result is to make the thing easier to use and to satisfy users’ actual needs, mission accomplished.
When Brevity Attacks
A prime way we simplify is to reduce verbiage. And sure, all things being equal, the fewer words or elements it takes to communicate something, the better. But all things are not always equal. Like making a room too cold, we sometimes overdo brevity, creating more problems than we solve. Better to use a little more verbiage to communicate clearly than to puzzle users with something concise but cryptic. Consider:
Concise action button verbiage is a particularly sacred cow. You can’t get much more concise than the standard “OK” and “Cancel.” The word “OK” itself is pristinely simple, but that simplicity comes at a price: additional effort and occasional grief. To know what you are consenting to in a dialog box, for example, you have to go to the trouble of reading the rest of the dialog. Users often don’t, so they sometimes suffer by clicking “OK” with regards to things they did not intend to. To spare them a bit of effort and grief, a self-explanatory button is more helpful than a concise button. That tends to mean a tad more verbiage.
Here’s an example of how our well-meaning insistence on bare-minimum button wording can sometimes do more harm than good:
Making the buttons self-explanatory (though less concise) can actually streamline the whole dialogue:
Is this not simpler than the original?
Bottom line: Simplicity is good. Ease is better. Brevity is good. Clarity is better. What are some of your favorite examples of clarity and ease in design? Share your examples in the comments below.
Image of single cell organisms courtesy Shutterstock.
Comments(11)
ABOUT THE AUTHOR(S)
John Boykin answers to user experience designer, information architect, interaction designer, and information designer. He’s been doing all of the above for over 10 years, sometimes as staff, usually as a consultant. Clients include Walmart, Macy’s and Bloomingdale’s, Blue Shield, Bank of America, Visa, Symantec, NBC, HP, Janus, and Mitsubishi Motors. His top-rated 3-minute video about user-centered design is used by many UX departments to evangelize UX to their clients and within their own organizations.
Monday, May 12, 2014
After Editorially: The Search For Alternative Collaborative Online Writing Tools
via Smashing Magazine http://ift.tt/1jNmf5K
With no means of annotating edits within a document, editors have to resort to surrounding their comments with square brackets. (Large preview)
Poetica has an incredibly polished and visually delightful interface, yet its utility may be limited to proofreading smaller texts. (Large preview)
Draft allows authors to review changes using an interface familiar to programmers — but is it well-suited to editors? (Large preview)
Typewrite has a beautifully stripped down interface, although its Markdown editor is a little buggy. (Large preview)
I’m going to let you in on a little secret: the best writers, be it your favorite authors or those that write for Smashing Magazine, don’t do it alone. Often, they work with an editor (or two), who will help them coalesce their words into something more compelling or easier to understand.
Having worked with several editors — and having been a technical editor myself — I’ve really come to appreciate this aspect of the writing process. Refinement is an essential aspect of any creative process. As refactoring code can make a program more logical and efficient, editing a text can allow an underlying idea to be more clearly stated, or make a piece more enjoyable to read.
I’m always on the look out for tools that can help me improve my craft. When I heard aboutEditorially, a collaborative writing tool designed with Web writers in mind, I immediately signed up. It quickly became an indispensable tool. From blog posts to talk outlines, I was bugging friends to provide feedback on my documents, which Editorially made very easy for them to do.
It had its shortcomings, of course. Two people couldn’t edit a document at the same time, while its focus on Web-based writing curtailed its utility; trying to tech edit a book with multiple chapters pushed the app beyond its natural limits. With the announcement that the service will close at the end of May, I’m left looking for an alternative. Can anything fill the void left by its untimely demise?
What Makes A Good Online Editing Tool?
Before we explore the alternatives, let’s examine what makes for a good Web-based collaborative writing tool.
- Distraction-free
Writing can be a difficult task, with writers prone to bouts of extended procrastination. In the space of writing this article I’ve cleaned my flat, washed the dishes and ironed clothes I don’t even wear! That is to say, the fewer distractions a writing tool can offer, the better. - Markdown support
Markdown is popular with writers precisely because it requires little interaction with a UI, allowing them to focus on their words instead. Any Markdown editor should therefore provide keyboard shortcuts (bold, italic, heading levels, and so on), with the resulting text appropriately rendered; if you’ve wrapped a word in single asterisks to denote emphasis, you’d expect it to be appear italicized. Text also needs to be easy to read. Font face, type size, leading, and measure should be chosen with care. - Annotation and discussion
Writing is just half the story, and a true collaborative writing tool should aid the activities that follow: sharing, editing, discussion, and review. Editors, co-authors and other collaborators should be able to highlight or replace passages of text, and annotate these changes if required. When working with multiple contributors, everyone’s remarks should be easy to differentiate. Somewhere to hold a general discussion about the document is useful, too. - Document handling
A document can go through several revisions, so versioning is important to ensure earlier thoughts can be reinstated or referred back to later. Writers and editors can be working on many documents at a time, so having a means of organizing files and indicating their status (draft, review, final, and so on) is essential. - Import and export
Finally, any tool should fit in with various production workflows. That means lots of import options (email, Dropbox, upload, sanitized copy-and-paste from Word) and plenty of export options (plain text, PDF, HTML, and the rest).
The Contenders
With these requirements in mind, what products are available, and how do they compare? When I heard about the closure of Editorially, I asked on Twitter for some recommendations. Here are some of the suggestions I received:
GITHUB
One of the first suggestions was GitHub. While it is by no means designed for writing documents, for many writers on the Web, this is where their content will end up. Any text file can be edited directly within GitHub’s website, although the interface is more attuned to writing code than prose. Being based on Git, versioning comes free, although managing branches and pull requests may be a little convoluted.
PROSE
If working directly with files in a GitHub repository appeals, you can improve the writing experience by using Prose, a beautifully designed and open-source editor built by Development Seed. Once you’ve authenticated with GitHub, the interface sits on top of your repos, providing an experience more suited to writing. This can be customized to expose any custom fields you have set up on a Jekyll or GitHub Pages install, too. Again, collaboration isn’t really a feature beyond that already provided by GitHub, but for small teams accustomed to that workflow, Prose may be a good fit.
PENFLIP
Taking the Git analogy a step further, Loren Burton created Penflip, which he describes as a “GitHub for writers“. If you’ve used GitHub, you will be immediately familiar with this tool. Much like GitHub repos, documents (termed projects in Penflip) are viewable publicly by default, but if you wish to make them private, you can pay to do so (prices currently start from $8 a month).
The editing interface uses the same one as open-sourced by Prose. This is a great decision, as Development Seed has created a fantastic editor that provides keyboard shortcuts and fully styled Markdown that is essential in such a tool.
With no means of annotating edits within a document, editors have to resort to surrounding their comments with square brackets. (Large preview)
The editing experience is a little more complicated. Once a document is shared with another Penflip user, they will then have their own copy to which they can make edits. These changes are then submitted back to the master version, which the original author can accept, comment on, or ignore. This enforced Git-like workflow prevents more piecemeal collaboration, however, instead preferring an all-or-nothing affair with regards to editing.
POETICA
OK, let’s move on from Git-based workflows, and try something completely different. Poetica, currently in private beta, has an incredibly innovative (dare I say, skeuomorphic) interface that uses a written proofreading notation that copyeditors will be familiar with.
From a technical standpoint it’s a marvel to look at, but in practice it can be a bit finicky to use. In terms of the features outlined in the introduction, Poetica lacks in many areas. There is no support for Markdown, no place to discuss changes, while heavy edits may overwhelm the text being reviewed (in a few cases, remarks overlapped one another).
Poetica has an incredibly polished and visually delightful interface, yet its utility may be limited to proofreading smaller texts. (Large preview)
Poetica is clearly a very proficient tool, but I suspect its focus is much tighter than the other tools covered in this article. As a proofreading tool it could be really useful, but for longer works and more complex editing, writers may soon find its limits.
DRAFT
A popular suggestion was Draft, developed by Nathan Kontny. It’s easy to see why, given its uncluttered interface and a slew of features, including a “Hemingway mode”, the ability to send text to a simplification service, and the option to have documents copyedited by “college-educated staff” (15 minutes costs $8).
So much of what makes a good writing experience comes down to the details, though, and I can’t say I enjoyed using Draft. The distraction-free interface comes at the cost of lengthy menus and model windows, with the interface lacking consistency in these areas. You can customize the typeface and font size used within the editor; that said, I never felt the need to alter this in other apps, but I did here.
Draft allows authors to review changes using an interface familiar to programmers — but is it well-suited to editors? (Large preview)
When it comes to editing, changes are reviewed with a diff-like interface that I found truly baffling. Again, we see a tendency towards using programming concepts ill-suited to the free-flowing collaboration seen between writers and editors.
TYPEWRITE
Typewrite aims to be “one of the best writing tools you’ve ever had”. It’s certainly one of the most beautiful, and competitors could learn a lot from Typewrite’s streamlined interface. Its Markdown editor is comparable to that of Prose and Penflip, although I found it to be quite buggy (inaccurate text selection and incorrect undos) but these issues can be easily ironed out. I also found it odd that the document title was missing from the UI, instead hidden behind a menu item. It’s the small things.
Typewrite has a beautifully stripped down interface, although its Markdown editor is a little buggy. (Large preview)
Perhaps Typewrite’s most interesting feature is its support for real-time editing, which allows multiple authors to work on the same text at the same time. This will be familiar to users of Google Docs (only excluded from this article due to its lack of Markdown support), although there is no means of knowing which author is making which changes. This feature also comes at the cost of any true editing features: there is no means of highlighting passages of text and adding remarks, for example.
As it currently stands, this tool is best suited to co-authoring rather than collaboration between an author and an editor or reviewer.
ONWORD
Also worthy of mention, Onword is a straightforward, stripped down editor created by Dan Eden. It features no collaboration or editing tools, and its Markdown editor is fairly basic. And yet, I find the simplicity of this app compelling. If Dan were to expand its capabilities to provide a workflow for editing documents, I would be a happy man.
And The Winner Is…?
Before I sat down to write this article, I had only briefly glimpsed the homepages of the alternatives people had suggested to me. Each is beautifully designed and full of promise, but you should never judge a book by its cover (pun intended).
Only when examined in detail, do you realize the Herculean task that lies ahead of anyone wishing to write such an app — and for an audience so attuned to the details. Having reviewed the alternatives, I realize how much I took for granted in the design of Editorially, for which the highest praise that can be given is: it stayed out of the way.
In Editorially, we had a product for editors, by editors. In surveying the landscape of its competitors, we see many that provide wonderful, easy to use and distraction-free writing interfaces, but fail to understand the editing process. Mapping programming metaphors like branching and diffs seems like great ideas in abstract, but in practice impede the less structured act of review and refinement.
It’s important to remember that many of these tools are built as side-projects, whereas Editorially benefited from a full-time team working on the product for a year. There is plenty of promise in each of these apps, and with continued development we may see a worthy successor. Until that time, I suspect many will revert to whichever hacks and broken processes they were using previously: word-processors, emails, printouts and scribbled pen marks.
(og, il)
Front page image credits: Torque.
Subscribe to:
Posts (Atom)