Autocorrection software, also referred to as text replacement software corrects misspellings and grammar as you type. Depending on the type of autocorrection software, these changes to your writing will either happen automatically or the software will suggest changes that you can make to a word or a phrase you have just written. Autocorrection is a standard feature on most smartphones and tablets because it is included in operating systems like iOS and Android and it is commonly integrated into text editors and word processors. Autocorrection, predictive text, and spell-check features have changed the way people type and edit their documents.
The algorithm that runs each kind of autocorrection software varies. In some cases, the algorithm may try to match the closest word to a misspelled word. Alternatively, the algorithm may suggest correctly spelled words based keys nearby on the keyboard to the word you have spelled incorrectly . Furthermore, some algorithms have the capability to adapt to your writing style in order to better predict what you are likely to say next. While autocorrection software continues to increase in accuracy, it does not always perfectly produce the words you have intended to type. The mismatch between user intent and results provided by autocorrection, predictive text, and spell-check often leads to comical or nonsensical pieces of text.
One benefit of autocorrect is the automatic nature of most autocorrection software. The software is able to classify misspelled words quickly so that writers can correct mistakes as they type. In addition to being helpful for anyone who types, this feature can be especially helpful for people with dyslexia. Rather than people having to manually spell-check their writing, autocorrection software can save people time by checking text as you type. Furthermore, autocorrect can help people by catching misspelled words that they might not have caught on their own .
There are some ethical critiques against autocorrection software. First, some wonder how potential biases in autocorrection software not only exclude people with certain identities but also how this software might reinforce cultural and racial biases in the way it decides what grammar and phrases are "correct." Additionally, others have questioned how predictive text impacts the way we communicate with other people. It has been argued that predictive text gives our communication a reductive and anti-individualistic quality. Some are concerned that predictive text leads to less meaningful thought and conversation.
The idea of autocorrect software stems from an ideology of "Do What I Mean" (DWIM) technology. DWIM technologies attempt to guess users' intentions rather than following users' input that may be incorrectly formatted . DWIM technologies and the idea of software correcting text as it is typed date back to the 1960s. It was during this time that a computer scientist created the idea of DWIM computer systems and invented the "undo" command .
In more recent history, the invention of autocorrect as we know it today is attributed to Dean Hachamovitch. Hachamovitch is a former vice president at Microsoft and first began working at the company in the 1990s. At the beginning of his career at Microsoft, he was placed on the Word team. At the time, Word already had a feature in which users could add shortcuts to their glossary. For example, someone could choose words like "insert logo" to be associated with their company's logo. Then, a user could conveniently type "inset logo" and press the F3 button to replace the text "insert logo" with an image of said logo. Drawing inspiration from this existing feature, Hachamovitch realized a similar mechanism could be used to correct commonly misspelled words .
Using a list of common errors people make while typing, Hachamovitch, and his coworkers developed software that would automatically correct these common errors while people were typing. These errors included mistakes like accidental capitalizations and common misspellings of words like "the." The replacement of these errors was triggered by the space bar, which is possible because words in English are always separated by a space character .
Types of Autocorrection and Spell-Checker Software
Despite its foundations in Microsoft Word, autocorrection software is integrated into many different kinds of applications. Word processors, smartphones, desktops, and browser extensions like Grammarly provide users with the tool to make their writing more efficient by suggesting words and by fixing spelling and grammar errors automatically.
Specifically, the AutoCorrect feature in Microsoft Office is advertised as a "feature to correct typos and misspelled words, as well as to insert symbols and other pieces of text." Furthermore, "AutoCorrect is set up by default with a list of typical misspellings and symbols, but you can modify the list that AutoCorrect uses" .
When a user is typing in a Microsoft Office application and makes a common mistake like typing "hte" instead of "the" the mistake will be automatically corrected once the user hits the space bar. Additionally, the spell checker feature will note words it does not recognize with a red, wavy underline to alert the user that they may have made a mistake.
In Microsoft Office, the AutoCorrect settings are found in the File tab under the Options section. There are five main AutoCorrect settings. The first of these options is the ability to customize AutoCorrect to automatically replace words with other words that you specify. Second, there is a setting called Correct TWo INitial CApitals. For every word you type, if you type the word with the first two letters capitalized and that word is in the dictionary, then the second letter is automatically made a lowercase letter. Furthermore, the Capitalize first letter of sentences setting recognizes the beginning of a new sentence and capitalizes the first letter of the first word of that sentence. Next, the Capitalize names of days option works similarly to the previously described functions and automatically capitalizes any day of the week that a user types. Finally, the Correct accidental use of cAPS LOCK key option is activated when a user types a word that begins with a lowercase letter, presses the Caps Locks key on their keyboard, and then types the rest of the word in all capital letters. This feature automatically corrects this mistake by making the first letter of the word capital, making the remaining letters lowercase, and turning off Caps Lock .
The Microsoft Word and Outlook programs within Microsoft offer two more AutoCorrect options: Capitalize first letter of table cells, which automatically capitalizes the first letter of the first word in a table cell, and Automatically use suggestions from the spelling checker, which prompts AutoCorrect to collaborate with the spelling checker feature if AutoCorrect does not recognize a word.
Users can disable and set exceptions for each of the AutoCorrect features in the AutoCorrect settings.
Google Docs, another word processor, also offers autocorrection features. Users can view, turn off, or customize these features by opening a new document in Google Docs, selecting Tools from the task bar, selecting Preferences from the drop-down menu, and selecting General. These features include: Automatically capitalize words, Use smart quotes, Automatically detect links, Automatically detect lists, Automatically correct spelling, Show Smart Reply suggestions, and Show link details. If a user select the Substitutions tab under Preferences rather than General, then users are able to create custom text substitutions.
When typing in Google Docs, if one of the autocorrect features automatically changes a word, that word will briefly have a gray, dotted underline. If a user hovers their mouse over the underlined word, then they are given the option to undo the automatic change with one click .
Apple describes its Auto-Correction feature as follows: "Auto-Correction uses your keyboard dictionary to spellcheck words as you type, automatically correcting misspelled words for you. To use it, just type in a text field" . Similar to the other kinds of autocorrection software described above, Apple's Auto-Correction software automatically replaces words it thinks a user has misspelled with the word it determines you meant to type.
Apple also offers a predictive text feature. This feature works by displaying three-word predictions above the keyboard as you type. This way, if the word a user is typing appears above the keyboard, the user can simply tap the word and it appears in the text the user is typing.
As with other kinds of autocorrection software, users have the option to customize text replacement. To do so, users can navigate to their settings application, go to the General section, then Keyboard, and finally Text Replacement.
Selecting the plus icon Add button prompts users to add their desired text-replacement phrase and shortcut. Users can also remove a text replacement by selecting the Remove button.
Finally, users do also have the option to turn off Auto-Correction, spell checking, and predictive text in the Keyboard section of their General settings .
Grammarly's mission is "to improve lives by improving communication" . The company was founded in 2009 by Max Lytvyn, Alex Shevchenko, and Dmytro Lider and their spelling and grammar checking product is available to users via Grammarly for Windows and Mac, browser extensions, mobile applications, and a web editor.
Grammarly currently offers both a free and a premium option for users. The free version of Grammarly is advertised to automatically suggest ways to help with spelling, grammar, punctuation, and conciseness. The free plan is recommended for what Grammarly refers to as casual writing. On the other hand, Grammarly recommends the premium version to help people with writing for work or school. The premium version is advertised to do everything the free version of Grammarly does plus provide recommendations related to clarity-focused sentence rewrites, tone adjustments, plagiarism detection, word choice, formality level, fluency, and other advanced suggestions.
When someone is using Grammarly, a red underline will appear under a word or series of words for which Grammarly has a suggestion. If the user hovers their mouse over the word(s) with the underline, they will see the change Grammarly has suggested .
Grammarly gives users some options to customize the way Grammarly works for them. To view and change these options, people can access their account on Grammarly's website and select the Customize option. Here, users are able to changes their settings related to: their personal dictionary, language preference, fluency suggestions, and editor settings. A users' editor settings include:
Additionally, you can configure your Editor settings when editing a document in the Grammarly Editor. This option allows you to: turn off auto-jumping to the next alert, preventing Grammarly from checking quoted text, and increasing the document’s font size .
The suggestions that autocorrection software makes, especially when users are sending messages to others via smartphones are widely regarded as a source of comedy in pop culture. For instance, it is common for people to write and send a text message without realizing that autocorrection software changed a word the user did not realize they had misspelled with a word the user did not intend to type. Generally, the comedic aspect of these mistaken text messages comes from changes the autocorrection software makes that transform a contextually appropriate text message into a vulgar text message. This phenomenon has resulted in the generation of articles that display mishaps in texting conversations due to autocorrection software. A simple Google search of "autocorrect fails" will lead you to such articles.
Even more, the idea of "autocorrect fails" has inspired a board game called Autocorrect. This board games includes 200 cards with photographs of fake text message conversation in which one of the messages is nonsensical due to errors that were clearly caused by autocorrection software. To play this game, two to four players each have their own dry-erase boards and race to write down what they think the send of the text message intended to type. Players win by correctly decoding the most incorrect text messages .
Autocorrection Software and Dyslexia
In 2014 Hiscox, Leonavičiūtė & Humby published a paper, The Effects of Automatic Spelling Correction Software on Understanding and Comprehension in Compensated Dyslexia: Improved Recall Following Dictation, about a study they had conducted at Cardiff University to test whether or not autocorrection software was beneficial for those with dyslexia. Specifically, this study aimed to determine the effects of the autocorrection software Global AutoCorrect, which has more advanced features than Microsoft's AutoCorrect, on recall and understanding of written content for students with dyslexia. Hiscox et al. explain that people with dyslexia have a working memory deficit which may explain the difficulty people with dyslexia have with writing because it is the working memory system that controls the many complex processes in our brains like text generation, transcription, attention, planning, reviewing, and revising, that give us the ability to write. This study did find an increase in recall and reading comprehension for students with dyslexia who used Global AutoCorrect. Based on the results of this study, researchers hypothesize that when students with dyslexia use software that will automatically correct their typographical errors, students can allot more of their working memory to the content they are typing rather than to the spelling of the individual words they are writing. Furthermore, the researchers think that regular use of autocorrection software by students with dyslexia may improve their writing and reading comprehension .
Based on this study, autocorrection software has benefits for the common good as it can be used as an assistive technology to help those with dyslexia worry less about spelling, giving them the space for their working memory to focus more on the content of their writing.
Racial and Cultural Biases
One common critique of artificial intelligence and predictive software is that an algorithm can perpetuate societal biases. For example, the case of the Amazon recruiting system that was found to be setting aside the applications of female candidates .
A similar critique has been made about autocorrection and spell-checker software. Rashmi Dyal-Chand, professor and researcher at Northwestern University published a paper in March of 2021 entitled Autocorrecting for Whiteness. Dyal-Chand writes about Microsoft Words AutoCorrect's biases towards names that are not traditionally Anglo or caucasian: "Across a range of products and applications, autocorrect consistently 'corrects' names that do not look White or Anglo. Sometimes autocorrect changes names to their closest Anglo approximations (as in Ayaan to Susan). Sometimes it suggests replacements that are not proper names (as in DaShawn to dash away). Often, autocorrect asserts the implausibility of non-Anglo names by underlining them in red". Dyal-Chand states that this phenomenon of marking names that are not traditionally Anglo as 'incorrect' is not simply a 'glitch' but is a feature that is disproportionately harmful to communities of color. Furthermore, Dyal-Chand argues that not only does autocorrect harm those without Anglo identities by enforcing whiteness as the default but simultaneously adds value to more privileged users. Dyal-Chand continues by proposing "design principles for ensuring more transparency, access, and participation in the design and deployment of autocorrect technology" .
In summary, Based Dyal-Chan’s argument, autocorrection software has negative ethical impacts because it labels names it does not recognize, which seems to be disproportionately names that are not traditionally Anglo. When software labels something like a name, which is often something largely influenced by an individual’s culture, as a word that is incorrect, this software is upholding the dangerous narrative that whiteness and Anglo culture are the default. Such biases in algorithms can be linked back to the datasets that these algorithms were given to learn from. Often, these datasets lack the kind of diversity that makes the resulting software completely accessible to all kinds of users. For instance, there have been many facial recognition software products that have clear tendencies to be better at recognizing individuals who are male and individuals with light skin rather than females or people with darker skin .
Recommender systems are integrated into software that people use every day. Social media, streaming services, and online stores use recommender systems to prompt users to watch, consume, or buy something based on users' previous actions on these online platforms. Some have wondered how being recommended what to type via predictive text systems impacts those who use autocorrection and predictive text software.
In addition to Auto-Correction and spell-checking, Apple also provides users with a feature called QuickType. Apple describes QuickType as software with the goal to predict “what you’re likely to say next. No matter whom you’re saying it to" . Furthermore, Apple claims QuickType is able to learn how to change its recommendations based on who you are conversing with on iMessage. For example, “your choice of words is likely more laid back with your spouse than with your boss” . Evan Selinger, an associate professor of philosophy at the Rochester Institute of Technology, speculates that predictive text features like QuickType could turn us into what he calls personalized cliches. Selinger explains: "Often without even realizing it, we’re all prone to lapsing into cliches when speaking and thinking quickly. When poetic license and creativity aren’t called for, it’s efficient – if not reassuring – to go for the cookie-cutter and conventional formulation. Cliches can be linguistic shortcuts that allow us to communicate without having to think too deeply about the words we use." Just as people rely on literary shortcuts when talking to communicate efficiently, Selinger wonders if predictive text features will also lead to more shallow thought and communication. Selinger warns "As communication becomes less of an intentional act, we give others more algorithm and less of ourselves... automation can be bad for us; it can stop us thinking" .
- “Autocorrect.” Autocorrect Definition, 12 Jan. 2018, https://techterms.com/definition/autocorrect. Retrieved 26 Jan. 2022.
- JobsForEditors. “Pros and Cons of Spell Checkers.” Editors, JobsForEditors, 21 Sept. 2020, https://jobsforeditors.com/blog/pros-and-cons-of-spell-checkers.html. Retrieved 26 Jan. 2022.
- “DWIM.” Wikipedia, Wikimedia Foundation, 16 Jan. 2022, https://en.wikipedia.org/wiki/DWIM. Retrieved 26 Jan. 2022.
- Engber, Daniel. “Who Made That Autocorrect?” The New York Times, The New York Times, 6 June 2014, https://www.nytimes.com/2014/06/08/magazine/who-made-that-autocorrect.html. Retrieved 26 Jan. 2022.
- Lewis-Kraus, Gideon. “The FASINATNG... Fascinating History of Autocorrect.” Wired, Conde Nast, 22 July 2014, https://www.wired.com/2014/07/history-of-autocorrect/. Retrieved 26 Jan. 2022.
- “Setting Autocorrect Options.” Microsoft Office 365, https://www.officetooltips.com/office_365/tips/setting_autocorrect_options.html. Retrieved 26 Jan. 2022.
- “Correct Your Spelling & Grammar in Google Docs - Computer - Docs Editors Help.” Google, Google, https://support.google.com/docs/answer/57859?hl=en&co=GENIE.Platform%3DDesktop#zippy=%2Cadd-a-word-to-your-spelling-suggestions-dictionary%2Cremove-a-word-from-your-dictionary. Retrieved 26 Jan. 2022.
- “How to Use Auto-Correction and Predictive Text on Your Iphone, IPad, or IPod Touch.” Apple Support, 27 Oct. 2021, https://support.apple.com/en-us/HT207525#:~:text=Auto%2DCorrection%20uses%20your%20keyboard,Tap%20General%20%3E%20Keyboard. Retrieved 26 Jan. 2022.
- “About Us.” Grammarly, https://www.grammarly.com/about. Retrieved 27 Jan. 2022.
- “Grammarly Support.” My Account and Grammarly Editor Settings - Grammarly Support, https://support.grammarly.com/hc/en-us. Retrieved 27 Jan. 2022.
- Stopera, Dave. “28 Autocorrect Fails That Spiraled Way, Way, Way out of Control in 2021.” BuzzFeed, BuzzFeed, 29 Dec. 2021, https://www.buzzfeed.com/daves4/funny-2021-autocorrects. Retrieved 28 Jan. 2022.
- “Autocorrect.” University Games, https://www.universitygames.com/games/party/autocorrect. Retrieved 10 Feb. 2022.
- Hiscox, Lucy, et al. “The Effects of Automatic Spelling Correction Software on Understanding and Comprehension in Compensated Dyslexia: Improved Recall Following Dictation.” Dyslexia, vol. 20, no. 3, 2014, pp. 208–224., https://orca.cardiff.ac.uk/61284/1/Hiscox%20et%20al%20-%202014.pdf.
- Dastin, Jeffrey. “Amazon Scraps Secret AI Recruiting Tool That Showed Bias against Women.” Reuters, Thomson Reuters, 10 Oct. 2018, https://www.reuters.com/article/us-amazon-com-jobs-automation-insight/amazon-scraps-secret-ai-recruiting-tool-that-showed-bias-against-women-idUSKCN1MK08G.
- Dyal-Chand, Rashmi. “Autocorrecting for Whiteness.” Boston University Law Review, vol. 101, 19 Jan. 2021, http://www.bu.edu/bulawreview/files/2021/03/DYAL-CHAND.pdf.
- Najibi, Alex. “Racial Discrimination in Face Recognition Technology.” Science in the News, 26 Oct. 2020, https://sitn.hms.harvard.edu/flash/2020/racial-discrimination-in-face-recognition-technology/. Retrieved 10 Feb. 2022.
- “Will Autocomplete Make You Too Predictable?” BBC Future, BBC, 15 Jan. 2015, https://www.bbc.com/future/article/20150115-is-autocorrect-making-you-boring. Retrieved 27 Jan. 2022.