8

Main Question:

I am a current undergraduate student who just attended my first conference and gave a podium talk on my research from last summer. I tried talking with a researcher who is well-known in the field about his views and he was rather dismissive. I am not sure how I should have handled the situation at the time or in reflection.

On one hand, his view is likely shared by only a small minority of people in the field, the view that no machine learning should be used in healthcare unless it is explainable. He proposes we should use methods such as symbolic logic AI instead (what he did most of his career).

On the other hand, I am just an undergraduate in the field and he is a very established researcher. I would expect him to be more correct than me.

1) How should I view the interaction? Should I assume I am probably wrong? Should I assume that he is wrong? Another option...?

2) In the future, should I press the debate more rather than only asking a few questions and thanking him for the info? Would this be impolite at a conference? Would it be an unwise career move?

Supplementary Info/Thoughts:

I find this balance of overconfidence vs underconfidence confusing in general for academics. As background, I am fairly successful relative to others my age. I have several first author publications, triple major, high gpa, and some awards. I was told by my adviser I would be a strong contender for the top PhD programs in my field. So regarding my placement relative to others my age, I view myself near the top, but not an outlier.

Placing myself and therefore my possible contributions regarding people older than me is more difficult. It is easier to place myself in formal sciences such as math where if I disagree with a senior research then there is a 99% chance I am wrong and they can prove it to me definitively. However, in less clear-cut fields such as health informatics, I am sometimes unconvinced by the arguments of a more senior researcher.

Should I raise my disagreement when this occurs? How long should I debate the issue if we don't agree? If we don't agree in the end should I assume I'm wrong?

I must sometimes come up with ideas that are both new and correct as evidence by my publications. However, given how much more likely a senior mathematician is to be correct than me in the event we disagree, I am not sure what to do/think when I disagree with senior researchers in fields such as health informatics.

Edit: If anyone would be interested in discussing symbolic AI vs ML in healthcare I would be glad to do it. I currently feel unresolved since I could not do it during the conference and would like to know what points I am failing to see.

  • no machine learning should be used in healthcare unless it is explainable, what does explainable mean? – user2768 Aug 28 '19 at 16:05
  • 1
    @user2768 Yes, like for the last question on the same theme, explainability and interpretability are indeed technical terms in the machine learning context. – Anyon Aug 28 '19 at 16:15
  • 1
    @xedg It's worth keeping in mind that this really is a limitation of current machine learning methods. It seems weird to single out healthcare applications though. Maybe you could have gotten an interesting reaction by asking if we should also refrain from using drugs if we don't know the mechanisms through which they work... – Anyon Aug 28 '19 at 16:32
  • 1
    @Anyon Perhaps the use of machine learning in healthcare can be governed by existing legal frameworks, such as those used for drugs – user2768 Aug 28 '19 at 17:08
  • 3
    In my experience this is not a minority opinion, and part of his dismissive stance may have stemmed from him having had this exact same discussion dozens of times before (due to current hype around deep learning). – xLeitix Aug 29 '19 at 8:20
4

1) How should I view the interaction? Should I assume I am probably wrong? Should I assume that he is wrong? Another option...?

Researchers should have differing opinions on topics such as whether non-explainable machine learning techniques should be used in healthcare.

2) In the future, should I press the debate more rather than only asking a few questions and thanking him for the info? Would this be impolite at a conference? Would it be an unwise career move?

Such opinions should be debated, that's how research advances, and such debate should be encouraged, at conferences and elsewhere. Some researchers will be less willing to accept your opinion than others, only you can decide whether you want to engage with such researchers.

Should I raise my disagreement when this occurs?

If you want to.

How long should I debate the issue if we don't agree?

As long as the conversation lasts.

If we don't agree in the end should I assume I'm wrong?

No. There doesn't seem to be any right and wrong here, just opinions.

| improve this answer | |
  • Could it be considered an opinion though if we share the same defined goal? In other words, we can disagree what the best metrics of performance are, but one method will probably perform better than the other by the given metrics. – xedg Aug 28 '19 at 17:34
  • I also was at first under the impression that the point of conferences was to improve research through discussion. However when I look at what the more senior researchers are doing, it seems more common for them to maybe ask a single question that hints at their disagreement and then leave it at that. My adviser was saying he thought several studies were not very good, but I did not see him talk to the presenters at all. – xedg Aug 28 '19 at 17:40
  • 1
    @xedg I think you've already touched upon what I meant by opinion, e.g., with your formal sciences remark. the point of conferences was to improve research through discussion, conferences have many points, it depends a little on the field. ask[ing] a single question that hints at their disagreement and then leave it at that is perhaps a tactic to confirm their suspicion. – user2768 Aug 29 '19 at 6:40
2

xedg, there seems to be two issues here which seems to complicate your mixed feelings. One issue is around status, which you alluded to, with your first author publications and strong academic record. The second issue is around where is the best environment to debate and clarify the pros and cons around the role of machine learning.

The first issue around status is quite unfair as you positioned the professor as being disrespectful or "dismissive". First conference presentations are usually always given great courtesy especially in question time. So what you interpreted as dismissive, may possibly be respectful distance. You are also not in a PhD program or have not finished your PhD, so experts will always back off and not put you in difficult spots out of respect of your position despite your accomplished publication record - common conference etiquette blog Remember having a bad experience at a conference will look bad for the conference and may also discourage people who may want to do a PhD. Attacking junior researchers will also diminish the professor in the eyes of other senior experts. Making junior presenters cry has been a subject of discussion among experts and does not look good at all. So all these factors may explain why debating an undergrad is poor form at a conference question time post-presentation. The other possible issue is around familiarity. As you attend more conferences and become a familiar face, people will be far more comfortable debating and clarifying complex issues with you compared to your first conference.

The second issue is what is the most appropriate format for this debate. If you want to flesh out "the role of machine learning in healthcare" at a conference, the short question time post-presentation is definitely not an appropriate timeslot for such a complex discussion. Symposiums or panel discussions are far more a satisfying format with the appropriate adjudication and moderation for this to occur. A senior mathematician would have a very different perspective than a clinician for example, but in a panel format, all the views would be respected rather than just your narrow focus on what is "right or correct" as implied in your question.

A review article examining the pros and cons of symbolic logic AI as the dominant approach in healthcare then be more appropriate in an extended article, rather than at a conference? To expect an expert to synthesize his academic focus and all the publications in a field is unproductive and not representative in a short question timeslot post-presentation.

I liked this Indian Hills presentation on how not to be argumentative might be helpful. Although you might not be overly argumentative, it always pays to move away from the "correct vs wrong" mentality in the academic realm. Most things are "more right" and "more wrong" depending on the scenario and circumstance.

| improve this answer | |
  • 1
    Sorry I may have been unclear. In the instance I talk about, I approached the senior researcher for clarification after the official question time had ended and people were conversing. How much discussion is appropriate at this point in the conference? – xedg Aug 29 '19 at 7:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.