My new favorite word: “Hedging”

Written by Becca Harrison

Dr. Lillian Lee, a professor of Computer Science at Cornell University, contacted me at the end of the summer to share her recently published study, Hedge detection as a lens on framing in the GMO debates: A position paper. She had out-of-the-blue read about my research interests on LinkedIn, and wanted to share this unique, interdisciplinary spin on my interests.
A researcher in the areas of natural language processing and information retrieval, Lee studies the ability of computers to use human language as a communication medium. She studies the empirical and theoretical problems that arise in the pursuit of this goal. She felt I would be interested in learning more about her research, which is incredibly welcoming as I would have never considered how my interests in agriculture, particularly genetic engineering, parallel that being done in engineering.
Lee’s paper puts a computational spin on my interests in the rhetoric surrounding the GMO debates. She focuses on the concept of hedging, defined as:

“expression of a tentativeness and possibility in communication, or, to put it another way, language corresponding to ‘the writer withholding full commitment to statements.’”

In this paper, Lee argues that more computationally-oriented research on problems involving framing is needed. For example, in the GMO debates, almost all pieces of scientific literature, popular journalism, and the like hold a distinct, framed position — whether the bias is intended, recognized, or not.
One of the best examples of framing in popular literature is Michael Pollan’s “The Omnivore’s Dilemma,” where he provides “data” and “evidence” of the big-business, economic drive behind food technology, and lack of humility in the industry. What Pollan fails to make clear, and why he is so convincing, is that he is not a scientist — he is a journalist with the charismatic ability to write in a convincing, at times unintended-yet-successful, frame.
There are social and scientific researchers and writers who also work within this frame. For example: bloggers!
According to Lee, understanding the ways in which participants in public discussions frame their arguments is vital to understanding the process of forming public opinion. For the purpose of furthering the use of computational linguistics in analyzing these positions, Lee’s team raises the following question:

In the controversy regarding the use of genetically-modified organisms (GMOs) in agriculture, do pro- and anti-GMO articles differ in whether they choose to adopt a “scientific” tone?

Prior work on the sociology & rhetoric of science suggests that “hedging” distinguishes popular-science text from text written by scientists (professional). Lee’s group proposes a detailed approach to studying whether hedge detection can be used to understand scientific framing (and thus framing in popular-science and journalism) in the GMO debates.
Unexpectedly, their preliminary analyses suggest that hedges occur less frequently in scientific discourse than in popular text, contradicting prior literature’s assertions.
According to the paper, the issue of framing is of great importance when trying to comprehend how public opinion is formed. Framing is described as occurring “when (often small) changes in the presentation of an issue or an event produce (sometimes large) changes of opinion.” Having the ability to understand this opens the doors to studying communication, especially in regards to reaching people in politics and media.
In media coverage of transgenic crops and the use of them in food, Lee asked if pro-GMO vs. anti-GMO articles differ — not solely in regards to respect to word choice, but in regards to chosen “scientific” discourse. For example, by inclusion of fewer emotionally-laden words and more words that indicate uncertainty. For example, phrases often found in popular and scientific literature, including, but not limited to: it seems that, implies thatwe wish to suggest …
This begs the questions, then: Who “hedges” more — pro or anti-GMO individuals? Can it exclusively be said that one type of media, be it popular journalism, blogging, Twitter, or even scientific journals or politics, tend to hedge more? It appears that Lee was unable to isolate one answer, but I’m sure everyone has an opinion. What “hedging” have you encountered in the GMO debate, and where do you think it occurs most or has the most impact? Do you think it is intentional? Are you a culprit of it?

Written by Guest Expert

Becca Harrison is currently pursuing her PhD in science & technology studies at Cornell University. She is particularly interested in how consumers view, communicate, and respond to technology used in food agriculture, and how such study can be used to influence effective policy, increasing accessibility of such food domestically and internationally.

Science & Society


  1. Great post! Thanks to Dr. Lee for this interesting research!
    I have noticed that the more hedging I see, the more scientific I automatically expect it to be. I still look at it with a critical eye, but the hedging does make me more comfortable I am with the content. The sweeping claims in works that don’t hedge make me very uncomfortable.
    I think both anti and pro GMO people use less hedging language than they should. That’s why I don’t like saying I’m pro-GMO – instead I’m pro science.
    While both camps do it, I do think the lack of hedging is different between pro and anti GMO folks. Pro GMO people tend to make claims that exaggerate while anti GMO people tend to make claims that are just unscientific. Contrast “we need GMOs to feed the world” with “GMOs cause autism”.
    Of course GMOs aren’t a silver bullet, we need many different methods to produce enough food for everyone – from drip irrigation to intercropping to breeding to biotechnology. The claim that “we need GMOs to feed the world” is a claim of exaggeration.
    Of course there is exactly zero evidence that GMOs cause autism. Even if one particular GM trait was shown to cause autism or other health concern, it wouldn’t mean all GM traits cause that health concern. The claim that “GMOs cause disease x” is patently false. At best, it is the result of terrible confusion. At worst, it is a bald-faced lie.
    Regardless of who makes claims of which type, it doesn’t help anyone to make statements that aren’t properly hedged.
    For example, one poorly stated (un-hedged) claim made by pro GMO people is that “GMOs reduce pesticides”. The truth behind the exaggerated claim is that use of the Bt trait can reduce use of insecticides, and that the glyphosate resistance trait can facilitate a switch from more harmful herbicides to the relatively less harmful herbicide. Another way to say it is that the environmental impact of pesticides is reduced when using Bt and glyphosate resistant crops. The continued successful change in pesticide use hinges on proper use of the traits, which includes rotation for both traits and refuges for Bt.
    While the hedged claim is a success story that has been validated in the scientific literature, the un-hedged claim has resulted in some pretty silly rebuttals. Charles Benbrook has been saying for years that pro GMO people are wrong when they say “GMOs reduce pesticides” because he’s measuring by pounds of active ingredient! Of course no serious person would measure pesticides by the pound – we all know that pesticides vary widely in their environmental impact. By their lack of hedging, the pro GMO people are partly to blame for Benbrook’s silly claim (now, I can’t take claim for this idea, it was something Karl and I discussed a few days ago).

  2. Well said!
    I had this exact X vs Y comparison in a comment thread the other day and even used the same examples: “pro-GMO” people exaggerate evidence — in my case, it was exaggerating the likely cost impacts of prop 37 — while the “anti-GMO” people are outright lying (or are extremely misinformed) about the known health effects of GMO foods.
    But, of course, every time we label a person as “pro” or “anti” GMO, we’re playing right into a narrative that makes it hard for reasonable hedging to have influence. Sigh.

  3. Philip Tetlock has researched similar ideas. Basically, he’s found people who hedge (he calls them foxes) are more often correct than those who are positionally assertive (he calls these hedgehogs). The names seem reversed, but are taken from author/philosopher Isaiah Berlin.
    The basic Tetlock idea is here.
    A Long Now talk is here.
    He might have a Ted talk somewhere too.
    I think scientists tend to hedge more as they innately try to stem criticism from their peers. Science is also too full of counter examples of complete theory fails for participants get too cocky.
    Still, scientists can “frame” just as easily as anyone else. It is something we (I) should guard against.

  4. Agreed. For a long time, I always put pro and anti in quotes but there isn’t enough room on Twitter. Also it’s a pain when typing on my phone. The quotes are there in my mind’s eye. 🙁

  5. Anecdotal evidence I know, but I have found anti-GM posts on the web are far more likely to be written entirely in uppercase than pro-GM posts.
    This act is a sure indicaion for every right-minded human being that the said post contains absolutely nothing of any worth and should be immediately ignored.

  6. Well, we can parse the language and select our terms carefully. But it seems to me that what really matters is what people hear.
    There was an article I was commenting on in the Guardian about GM potatoes, the blight-resistant ones. I had parsed some language to make a statement.
    My statement: Well, it turns out that what likely brought the blight to Ireland was organic fertilizer. So I think that in fact we need to stop the organic farmers.
    The response:

    “Well, it turns out that what likely brought the blight to Ireland was organic fertilizer.”

    from that sentence its seems more like you are quessing ?
    What I said: likely. What he/she heard: guessing. Of course in my next comment I had to explain that we don’t time travel, and we can only based our claims on the evidence we have, and I delivered the evidence. But it didn’t matter.
    I’m really struggling to understand if the parsing or if the language analysis is telling us anything valuable. Does it matter to the readers/listeners? Does it really affect the discussion? Or could caution/hedging actually hurt? What’s effective?
    I’m not convinced that we have the data on this. And I’m also not convinced that all situations are the same and that any one strategy should be the focus.
    It’s not clear to me from the post whether Pollan is hedging or not. But either way–he’s apparently successful. Should we adopt that because it’s effective?

  7. The fact that labeling a person “pro” vs. “anti” GMO is us playing right into the narrative is so true — it’s unfortunate the debate has become SO polarized that there isn’t much of a middle ground. I suppose this is further exacerbated by simply putting the decision of labeling on a ballot. Anastasia, I am going to steal your idea and identify as a “pro science!”

  8. I never really thought about this, but it is certainly a valid point! I suppose I will keep my eyes open for this and see if it applies to non-science everywhere!?

  9. I’m going to identify as “pro-science”, too, although you would also certainly label me as “anti-GMO”! As such, it’s very interesting to see how all of you “frame” your comments in a distinctly biased fashion.
    An exaggeration, for instance, is simply one type of lie. Saying someone is exaggerating, however, is much less “emotionally-laden” than calling someone a liar, and so you chose to call “pro-GMO” people exaggerators and “anti-GMO” liars. It would be just as accurate to say that pro-GMO (dropping the quotes here, but leave them in your mind’s eye, please) people lie about feeding the world and reducing pesticides and the high costs of labeling, but you’d rather soft-pedal those as “exaggerations.”
    Then, of course, we can look at the examples chosen. Sure, let’s contrast the oft-heard, relatively-moderate claim about feeding the world (7.16 million Google hits) with the uncommon and outlandish claim about autism (156,000 results). Anastasia certainly could have found more comparable statements, but those wouldn’t have framed the argument in favour of her bias, would they have?!
    Looks like we can’t even have a discussion about hedging and framing without falling victim to it. Sigh…

  10. For instance someone might be exaggerating when they state that the pro-GMO arguement that GMOs will feed the world is oft heard based on 7.16 million google hits, my google-fu may be weak, but I return ~2million hits, and the first page is 80% anti-GMO sites/articles claiming that the claim is made.
    GMOs cause autism returns 157,000 results, 90% of which (on the front page at least) are actually making the claim.
    It’d be rather rude to call them a liar though.

  11. I am sympathetic to your ideas here. We all have frames whether we like it or not. But there is a fundamental difference between completely making up a falsehood and exaggerating. They are not the same at all. It’s the difference between “I caught a fish this big” and “I saw a mermaid”.

  12. I would guess (yes, I’m guessing) that it depends on which words you use to express your hedging. When you use the word “likely” as a scientist, you mean that your claim is supported by evidence. But “likely” also has a colloquial meaning, and outside of science, many people use that word to indicate that they are guessing. Perhaps your meaning would have better been conveyed by saying something like “Based on the available evidence, the most likely explanation is that…”

  13. Actually, I don’t think it would be rude to call someone a liar based on the inability to replicate an unspecified Google search (I did it again just now and came up with different numbers). Terms like “silly” and “completely missing the point” seem more appropriate, although likely rude!
    The point is that if I wanted to I could probably dredge up a rare and outlandish pro-GMO claim and compare it against a common and moderate anti-GMO claim to illustrate that one side “exaggerates” while the other completely makes up unscientific, patently false, silly, bald-faced lies (I apologize if I missed any of the colourful terms you’ve used to describe the anti- side!). And given my “frame” I probably would! 🙂

  14. The problem of course is the common and outlandish anti-GMO claims, they pop up all the time. You will rarely see a discussion of any size on GMOs where folk will not bring up unscientific, patently false, silly and bald faced lies (if you can point me to somewhere this isn’t the case that’d be dandy, I point you above as a for instance, there’s this one guy who rightly suggests that the feeding the world arguement could be categorized as a lie (we no more need GMOs to feed the world than we need better corn hybrids to feed the world, or better storage facilities in devloping nations to feed the world – they could all help, but none are a silver bullet), but who also suggests that the cost of labelling argument is a lie (even @ the conservative 60-70c (approximately what was recently put forward on Dr.Oz, so is most likely a total fabrication, but one by someone trying to downplay the cost) a consumer this is hardly small change (311 million people * 60c = $186.6M, or if only california (there wasn’t much clarity about that) 37M * 60c = $22.2M) or that the reduction in pesticides arguement is a lie (which is true only if you really really squint hard and refuse to consider things like EIQ while also blatantly ignoring the data)

  15. Hedging may be proper scientific behavior, but you can detect it in papers with little scientific merit. For example I have pointed out that Samsel and Seneff’s paper on the dangers of Roundup ( is full of what I call “academic weasel words,” which take the place of actual assertions. For example:
    One can surmise that
    Could be a contributing factor
    We develop a novel hypothesis
    We suspect this has to do with glyphosate’s effect…
    We hypothesize that
    …and this could be a factor…
    …could substantially enhance…
    It is plausible that…
    Is like synergistic in combination with glyphosate
    This effect can conceivable explain…
    We hypothesize that DHEA sulfate levels are a hormonal signal
    …as might be induced by glyphosate’s interference with tryptophan synthese…
    …can lead to inflammatory bowel disease…
    …could be anticipated that…
    These words mask the fact that the authors didn’t actually do any research at all.

Comments are closed.