“Don’t you see that the whole aim of Newspeak is to narrow the range of thought? In the end we shall make thought crime literally impossible…Every concept that can ever be needed will be expressed by exactly one word, with its meaning rigidly defined and all its subsidiary meanings rubb out and forgotten.” - from George Orwell’s 1984.

What is the most important book on the bookshelf? The answer is the dictionary, for it is the database of words and their meaning. Words are power, for in manipulation of their definitions exist the opportunity to mold them to whatever one wants them to be, thereby changing the reality of things. This is most apparent in the book 1984 by George Orwell where Winston’s coworker Syme, expresses excitement for his job in taking Oldspeak and converting it to Newspeak. Syme states his job in building the Newspeak dictionary is the active deletion of words, thereby changing their meaning - giving them multiple and contradictory meanings leading to doublethink. Doublethink takes a word with contradictory definitions (because we have fewer words to define things by) and uses whichever is convenient for the desired result. One can weaponize a word, where the word now means something that it does not to convince and persuade. The destruction of words and their meanings reduces the need for conscious thought, discussion or debate become unnecessary - discourse becomes untenable. Therein lies the power of the printed dictionary and why it is the most important book on the bookshelf.

“Doublethink means the power of holding two contradictory beliefs in one’s mind simultaneously and accepting both of them. To tell deliberate lies while genuinely believing in them, to forget any fact that has become inconvenient, and then, when it becomes necessary again, to draw it back from oblivion for just so long as it is needed…” - from George Orwell’s 1984.

The power of a printed dictionary is a fixed record of words and their definitions, immutable. Whereas, an online dictionary lacks permanence, able to be rewritten on the fly to suit shifting narratives or agendas. In an era where words can be weaponized, a printed dictionary is an unalterable reference point acting as a bulwark against the erasure of the past. On my bookshelf, I keep a copy of the New Oxford American Dictionary, preferring its hybrid blend of American and English words better fitting that of a writer or professional as compared to the Merriam-Webster or Oxford English. I would compare the three dictionaries with this analogy. The Merriam-Webster is USA Today, New Oxford American is the Wall Street Journal, and the Oxford English is the Financial Times.

**

Why discuss this? Because shared definitions to include the context and semantics of words across teams, divisions and organizations is hard - technology makes it harder. First, words are now more corruptable due to digital sources (as opposed to printed dictionaries) and they can be altered in the blink of an eye. Second, Generative AI, has the power to completely remap the systems, protocols, policies and procedures of an organization. The rapid automation of tasks, interactions with customers, clients and coworkers through the interpretive effects of AI raises concerns.

For words are more than just definitions, words are also about context and semantics. The semantics are the logical aspects of meaning such as sense, reference, and implications; while, context is what immediately precedes or follows to give the word meaning (Lindberg, Stevenson, 1587, 375). Finnish comedian Ismo illustrates this in his skit about the word ass rather well as he talks through semantics and context. Like, adding ass to the word bad changes the entire meaning of the word.

On the surface, this just sounds like understanding slang, but in the context of the organizational framework (systems, processes, procotols, procedures) and the specific words used to define this framework - misalignment in defition can change the context and semantics as well which can trickle down into the individual employees and customers. Because words matter; a weak foundation in the words, their context and semantics can distort the organizations messaging. This messaging can lead to internal confusion, operating at cross-purposes across divisions and teams (especially when there are physical boundaries such as geography between them). Further, it can influence the customer and their perception of the brand and its product.

To illustrate, I will highlight some common terms thrown about in any number of busines magazine articles discussing organizational behavior & development. These articles throw around terms like “systems”, “protocols” or “processes” to describe the novel method to create value through better teamwork. The initial issue is what I have already highlighted - words matter. The author may have different definitions. Along with different definitions is the terms may be taken out of the context & semantics to how those same terms are used in your organization. Worse still if the author conflates the terms and uses them interchangeably to make it more murky. Essentially, do these terms directly translate? Will implementing the author’s solutions work if there is not shared meaning? Perhaps, this analogy will help.

In the movie Office Space, the Initech principles are the company’s tenets (paradigm) and way of thinking. A principle is a fundamental truth or proposition that serves as the foundation for a system of belief, behavior or for a chain of reasoning (1388). Initech, is a software company, its theme is contained in its name based on the words initiative and technology (you could substitute innovative as well). Initech at its core is about lines of code, which requires buy-in by its employees into the Initech system, the injection of the company’s tenets in the execution of innovative code writing. A system is a set of principles or procedures according to which something is done; an organized scheme or method (1763). We see this in banners hung around the office stating, “Is it good for the company” and reminders that Initech is about “Initiative” combined with “technology” that broadcast the system that Initech is on the forefront of technology through the proactive efforts of its employees.

When Peter Gibbons, Samir Nayeenanajar (Samir N-Eleven) and Michael Bolton are working on the quality assurance of the company code, mainly to ensure it is Y2K-compliant, they do so within the protocols of Initech work environment. A protocol is an accepted or established code of procedures or behavior in any group, organization, or situation (1404). When Peter is promoted and decides to show up late to work and dismantles his cubicle wall and tear down the “Is it good for the company” sign - he is rebelling against the company’s protocols. Yes, there is similarity between system and protocol - I think of it as the system is the overstory and the subplot is the protocol.

Peter Drill Peter demonstrates destruction of Initech systems

When Bill Lumbergh asks Peter Gibbons where his Test Procedure Specification (TPS) cover sheet is, he’s referencing the company policy distributed to all employees. The Initech company policy is that all employees when submitting a TPS report must include a cover sheet. A policy is a course or principle of action adopted or proposed by a government, party, business or individual (1353). The series of actions performed when submitting a TPS report is to attach a cover sheet to it - this is the company procedure. A procedure is a series of actions conducted in a certain order or manner; established or official way of doing something (1392). Almost interchangeable with processes which is a series of steps to achieve a particular end (1392). It’s a subtle nuance in the definition of words and a clear example of how easy it is to quickly move away from shared definitions in an organization, particularly one with several departments and divisions (or silos). If the definitions of these terms are not shared, it becomes difficult for individuals and teams to understand their meaning and relationships - ultimately leading to barriers to integration. Thus, the importance of dictionaries.

What if we introduce Generative AI into the mix? Organizations need to think critically about the lexicon of words they input, their context, and intended semantics. Illustrating the existing difficulties of shared meaning, now imagine how integrating Generative AI increases the difficulty in keeping the definitions and context to key words and phrases shared across an organization and the external stakeholders (to include customers).

When speaking of semantics, this goes beyond just the definition of the word but how that word conveys meaning depending on the context, and without shared meaning we will have issues in context. These issues can balloon due to the penetrative manner in which AI can be integrated into the organization across all divisions to include customer facing engagements. Worse outcome, internally the company is charging forward based on a set of definitions that are completely different from what is being broadcast to the customers and in-turn what the customers think the semantics of those terms are. Imagine if the word “privacy” had different context and semantic meaning inside versus outside the organization? Look at the backlash against Apple over a difference of opinion of what “privacy” meant. Organizations are going to charge ahead and fully integrating AI despite the pitfalls. Why?

Efficiency. The hope of outsized efficiency gains drives organizations to embed AI into their systems. Turning employees loose on AI in research and development of products and service, or in creating digital prototypes to put in front of customers. Employees using AI to create advertising or analyzing past campaign performance and curating customer feedback. AI is augmenting and automating processes and procedures that use to be done manually by individuals. It’s not just about getting employees to buy into the company culture anymore, it’s AI as well.

This is a bit of a sidebar:

Word Vectors clustering of vectors

Large language models (LLMs) are really about statistics. LLMs turn words into numbers and then use statistics to predict the next word in the sequence. To do this, LLMs turn words into vectors, the more vectors, the more context, the more the LLM can “predict” the next word in the sequence. This ability greater depends on the transformer, not the 1980’s variety, the transformer architecture which allows the LLM to handle longer (larger) word sequences, long enough to gain context. The statistical predictive capability depends on words and tokens which are represented as vectors.

The more these vectors cluster in dimensional space, “bank”-“shot”-“court”-“hoops” - the more semantic similarity in vector space supports prediction. So, in this case, when using words as input into an LLM, words very much matter as word selection leads to vector representations that cluster away from the intended context and thus desired meaning. For example, offer up the words “bank” and “shot” and the LLM could interpret this as either a bank robbery or basketball game. It needs additional data (words) to derive the context of the words. More context adds “court” and “hoops” and now the LLM knows how to respond with words associated with basketball rather than a bank robbery. All of this requires learning and fine-tuning, further refined by human feedback (when ChatGPT asks you which of the two answers you prefer) to improve the statistical prediction capability of the LLM. Something that will be harder to accomplish without a foundational framework of shared definitions for the key words and phrases across the organization.

Again, the value of the dictionary and the tediousness of trying to be precise and truly believing in the mantra, words matter. One definition lead to another, leads to another - to truly know the meaning of an idea one must research many associated things and their meanings. This process does not lend itself well to rapid fire exchanges between individuals in organizations where the potential that shared meaning does not exist. Differences in bias, perspectives and context of problems wrapped in disparity in the definitions of words prevent shared meaning and ultimately conflict interfering with organizations ability to build and employ effective teams to solve problems.

^Employees that have job roles that deal with relationships, communications, advising, liaison….these roles are the conduit for shared meaning across organizations (both internal and external). They should be paid well. They likely are not because their role lacks metrics to gauge their value against other roles where there are clear quantifiable metrics to analyze job performance. This is a foul, organizational leaders need to recognize the intangebles and value them accordingly. It is not always about tangible physical output or hours worked - it is about impact.

__________

Judge, Mike. dir. 1999. Office Space. 20^th^ Century Fox, 2018, Blue-ray Disc.

Lindberg, Christine, Stevenson, Angus, ed. 2010. The New Oxford American Dictionary, 3rd ed. New York: Oxford University Press.

Mims, Christopher. 2025. “How the Owner of Hidden Valley Ranch Learned to Love ChatGPT.” The Wall Street Journal, July 4 2025. https://www.wsj.com/tech/ai/clorox-ai-hidden-valley-ranch-e997d3dc?st=Gc7rXF&reflink=desktopwebshare_permalink

Orwell, George. 1977. 1984. New York: Signet Classics.

__________

© 2025 Jeremy Reynolds, all rights reserved.

Back to Blog Index