Dictating in Third Person: Ends the I Think Problem

The draft arrives with a recognizable sound: I think the evidence points this way. It seems to me that the data supports the case. The sentences are grammatically fine, the ideas are plausible, and the essay as a whole never quite argues. The problem is not the presence of a voice; it is the constant retreat from it. Academic writing asks the writer to assert claims as propositions and to defend them in the body, and the I think hedge is a verbal stepping back at every crucial point. Dictating in third person is the most direct fix, because it removes the speaker from the sentence and leaves the claim standing alone where it has to be judged.

The deeper cause of the hedged draft is the assignment of a new persona. The student has written personal essays in the past, where the first person is the subject, and now writes about other people's work, where the argument should stand on evidence rather than personality. The I think is an attempt to be polite, to keep the self visible, and the result is a paper that reads like a series of opinions. This is a crowd-pleasing posture in conversation and a failure posture in an academic paper. The discipline being learned is the one that matters most: claims must be strong enough to be evaluated, and the first-person hedge is exactly the strength that is being sacrificed.

There is a fine line to notice: not every I is forbidden. Some assignments ask for a first-person voice, and some academic writing uses the first person comfortably. The specific problem is the hedge, not the pronoun. I think the model fails is different from I argue that the model fails, and the difference is the verb, not the person. The fix is therefore not a blanket ban on the pronoun but a ban on the hedge verbs, think, believe, feel, seem, and their kin, and an insistence that whenever they appear, they be replaced by the claim's own force. That force is the academic voice.

Dictation leaks the hedge at three predictable points, and naming them is the first step to closing them. The first is the opening of a paragraph, where the writer says I think as a warm-up before the actual thought, the way a speaker clears the throat. The claim follows a second later, and the hedge is doomed to stay on the page. The second is at the moment of transition, when the paragraph shifts from summary to analysis; the writer is about to take a position, and the hedge signals the risk. The third is the strongest: the moment the writer asserts something they are not sure is true, and the hedge is a plea for mercy. Each leak has its own cure, and the cure is the same: say the claim by itself, without the prefix.

The leak at the opening is the easiest to fix because it is simply deletion. Dictate the thought, then look for an I think, or I believe, or I feel like at the start of the paragraph, and delete it when you revise. The transition leak is harder, because it is structural: the writer is unsure whether they have the right to assert. The sentence that follows the hedge is usually the strongest sentence in the paragraph, and the fix is to trust it. The strongest leak is the third, and the fix there is research: a claim that needs an I think before it is a claim that needs evidence before it. The recording is the pointer, because the voice at that moment is apologetic even before the word is spoken.

The leak has a guest, the hedge words that accompany it: maybe, possibly, somewhat, sort of, and a bit. These are the same retreat at the level of the adverb, and they show up in dictated drafts at the same frequency as the I think, because spoken language is hedged by nature, and a person talking about an idea to a recorder sounds like a person talking about an idea to a friend. The next stage is to remember that a dictated draft is conversation, and a conversation is not a manuscript. The transformation is the task of the paragraph's frame: interpret, state, support, and it is a task that training makes automatic.

The discipline is one small change in the mouth: when the thought is ready, say it as a fact about the world, not as a fact about you. Instead of I think the evidence shows a trend, say the evidence shows a trend. The claim is the same, and the weight is different, because a statement can be tested and an opinion can only be shared. The training is basically a kind of rehearsal: dictate the claim as a news anchor would read it, flat and true, and the draft arrives in the register the assignment wants. This is not a trick of grammar. It is a change in stance.

The stance is supported by the paragraph's work. A claims-first paragraph announces its position in the first sentence, then supports it with evidence, then shows what the evidence means. The hedged version spends two sentences expressing uncertainty before reaching the claim, and the reader learns to discount the sentence exactly as heavily as the hedge said to. The claim-first version is not more arrogant; it is more testable, which is what an academic reader wants, and the test is the exchange: the reader can agree or disagree with a claim, but with an opinion there is nothing to test. The ear test is one question: could the next sentence contradict this one? If it cannot, the sentence is a hedge, and the paragraph has said what the evidence is, not what it means.

The practice has a naming trick. Say assert aloud, a single word, before each claim you dictate, and then say the claim. The word is a spoken marker like the punctuation vocabulary, but here it marks the mode instead of the grammar. The first few times, the word forces you to finish the sentence without the hedge; within a week, the marker becomes a habit and the mode stays on for the whole paragraph. The final effect is the one that matters to the reader: the essay reads as if the writer had looked at the evidence and made up their mind, which is exactly what a good essay is.

The comparison is a useful exercise. Dictated, a paragraph about media policy might read: I think the attention economy is a bigger problem than people realize, and I feel like the platforms should be held accountable for it. The rewrites say: the attention economy distorts public conversation more than its critics claim, and platform accountability should extend beyond content moderation to the design of the feed. The two say roughly the same thing, and the second gains in exactly the dimension the first lost: the second is arguable, testable, and ready to be supported by the body with evidence the first never calls for.

The rewrite has three moves, each teachable. First, delete the hedge and its subject. Second, let the claim stand as its own subject, which is usually the object that was being discussed. Third, add the evidence or the warrant that the claim now owes, for a claim that can be argued needs a reason, and a statement that can be tested needs support. The paragraph now does what a paragraph is supposed to do: it asserts, it explains, and it points to what it will prove. The order is subject, claim, connection to the next step. The result reads like a person who moves the field forward rather than a person who is unsure about the weather.

The exercise is worth doing on a full page at least once, and better twice: once to see how much of the essay was, in truth, commentary on the writer's own uncertainty, and once to see the effect of staying in the third person on the essay's final sentence. The second pass confirms the change. The paragraph that once said I think the model is limited now says the model's limitation is visible in its treatment of age, and the difference is felt in the rest of the paper. The revision is not about removing the writer; it is about moving the writer to where the writer belongs, in the analysis rather than in front of it.

The filter is cheapest when it is on from the first word of the first recording, and the way to turn it on is the threshold sentence described in the dictation process: state the register aloud before you begin. A small ritual changes the draft. The threshold sentence is a short statement: this paragraph will show that the design fails, and the sentence, and the claim, set the register. The benefit is that the threshold is also a claim, which, said out loud, anchors the paragraph to the task. The filter does not need to be perfect; it needs to be on, and the threshold keeps it running.

The filter has one more feature worth using: the catch command. If a hedge arrives mid-sentence, say hedge and correct the clause: the model, hedge, has a structural limitation. The catch is the same spoken correction used for tense and punctuation, and it trains the mouth to notice itself. After some weeks, the catch comes earlier, and after a term, almost none of the hedges arrive at all. The progress is not a matter of willpower; it is a matter of the auditory feedback loop, where the voice hears its own retreat and corrects it immediately. The recorder is the coach who never gets tired.

The final filter is the reading. Read your draft's claims out loud, one by one, and mark the ones you would be comfortable defending to a stranger. A claim you would defend is a claim that works. A claim you would qualify before the sentence ends is a claim that still needs the evidence or the narrowing it was trying to avoid. This is the test that makes the whole training real, because it connects the draft to the conversation the essay is entering. The I think problem is not a grammar problem at all; it is a confidence problem, and confidence is built by evidence, which is what the filter forces you to collect.