opinions-organizations-earlierTenure

1st DRAFT

A proposal to give tenure earlier

by Bayle Shanks[[image:who does not have tenure, in case you think that biases me?]]

ABSTRACT

An oft-heard complaint, particularly in interdisciplinary fields like neuroscience and artificial intelligence, is that lines of research which are risky, long-term, integrative, cross-disciplinary or speculative are not pursued as much as they should be. I postulate that the economic structure of academia is responsible for this. I propose that giving tenure earlier in research careers would be an effective way to fix the problem.



Contention 1: the economic structure of academia penalizes work that is risky, long-term, integrative, cross-disciplinary, or speculative

The evidence for these effects

Evidence for a deficit of integrative work in artifical intelligence:

Evidence for a deficit of integrative work in neuroscience:

i really should collect more evidence; please let me know if you have any that should be added

The causes of these effects

Long-term lines of research are penalized because researchers need to show results within a few years in order to continue to get funded or in order to continue to have a job.

Risky lines of research are penalized for two reasons.

First, an individual's economic destiny is tied to the success of their research, and there is economic value in reducing risk (you'd rather have a job that will definitely pay you a certain amount than a job that might fire you or might pay you twice as much, right?).

Second, the difference in value between the rewards for very successful research and marginally successful research (tenure and fame versus maybe tenure, and getting to keep trying) is much less than the difference in value between the reward for marginally successful research and the reward for unsuccessful, risky research (maybe tenure, and getting to keep trying, versus no chance of tenure and having trouble keeping a good job at all).

Integrative lines are research are penalized for three reasons.

First, all competitive economic systems skew reward not just towards progress but also towards more easily measured forms of progress. It's easier to see when someone has acquired valuable data or developed a valuable new technique, but the value of integrating stuff that was already known into a synthetic framework is harder to assess.

Second, if person A creates a synthetic framework, for this to be successful, you need everyone else to spend at least a modicum of time integrating their own specialized knowledge into this framework. But everyone else doesn't get credit for doing this; you get credit for your own initiatives, not for helping out with someone else's initiative. Someone who publishes lots of papers will get more credit than someone who publishes few papers but who gets tons of acknowledgements in other people's papers. This provides an economic motivation for "not invented here syndrome".

Third, since success in academia is assessed through publications, it is hard to reward someone for doing "more of the same". If someone creates an anatomical database, they can publish and get credit. If they spend the next five years editing the database, refining it, and writing new entries for it (but in the same domain as the previous entries), how do they get credit? Do they publish another article that just says "See my last article; I did that same thing for 5000 more entries"? If they do, will that second article earn them as much credit as the first?

Cross-disciplinary lines of research are penalized for two reasons.

First, no matter how much everyone says they like interdisciplinary research, the important economic decisions which affect an individual researcher's life are made mostly by departmental committees. For instance, tenure; if you convince three departments that you are 40% as good as other applicants, that won't get you a job, even though 3 times 40 > 100. Unless there happens to be a special cross-disciplinary position available at that university for exactly this intersection of disciplines -- but these are more rare than "normal" single-discipline spots; it's more risky to bet on getting one of those spots, and also your choice of which universities/institutes to work at is more restricted.

Second, people in each field will tend to disrespect anyone who doesn't "keep up" with the goings-on in that field. It does not impress people to say, "I don't know about that even though everyone else here does, but look, i don't need to, it's actually less related to my line of inquiry than this other stuff in this other field that you don't know about, that I do". This means there is essentially a fixed cost with each additional discipline that you want to publish in.

Speculative lines of work are penalized for two reasons.

First, and most importantly, speculative work is harder to evaluate, and hence as we discussed above, it is rewarded less in a competitive economic system.

Second, speculative work is less impressive-sounding. If someone says they developed such and such technique, you're like, woah, this guy really knows a lot more than me about this. If someone says they sat around and racked their brain and thought and came up with a better language for discussing what other people have done, you may think, that's neat, but I could have done that too if I'd spent my time sitting around instead of staying late in lab doing "real work".

Maybe you couldn't have come up with that; maybe you would have sat around and not come up with something as novel or creative or useful as that guy did; but often useful speculative insights sound "obvious", once someone else says it.

Contention 2: Therefore, academia could benefit from reducing competitiveness

All of the harms above derive from the competitive economic nature of academia. Worse, many of them are inherent to any competitive economic system, since competitive economic systems tend to reward:

The solution is to make academia less competitive. This is one of the primary advantages of the tenure system; after getting tenure, the impact of competition on an individual's personal finances is lessened. The researcher becomes correspondingly more free to pursue risky, long-term, integrative, cross-displinary, or speculative lines of inquiry.

Proposal: Give tenure earlier

One way to have "more tenure" is by giving tenure earlier in a researcher's career. Out of all of the people working in academia at a given time, this would increase the proportion that are tenured.

Evaluation of the proposal

If a little tenure is good, will more be better? We have to compare the upsides and the downsides

Advantages of earlier tenure

The first five of these have been discussed above, the rest require some explanation.

Iconoclastic research is research that an minority of researchers believe is valuable but which a majority believes is a waste of time. In the history of science, there are many examples of research that was derided as a waste of time when it was being conducted, but which was ultimately immensely valuable.

When one is permitted to work on what one really wants to work on, one does more and better work. Therefore, the younger a researcher gets tenure, the more time they will get to spend on the topics they really want, and therefore the field as a whole will have more passion applied to it.

Younger researchers. Younger researchers are often energetic and have put in many hours at work, on average more than they do when they are older. The primary reason for this is of course probably that they don't have tenure, so they are compelled to work longer hours, whereas some of the older researchers have tenure and so on average have to work less hard. However, I suspect there are other reasons. Younger researchers are less likely to have children, and raising a family takes a lot of time. Younger researchers often are energized about "getting started" and about "making an impact". And younger researchers haven't worked so many hours for so many years yet, and so may be less "burned-out".

Although I don't have data to support this, personally, I've noticed that a disproportionate number of postdocs and assistant faculty members tend to be very energetic and passionate about their work. This is not to say that there are not many professors who are very energetic or who work very long hours; I see many examples of that too. It just seems that the incidence of this is slightly greater in the younger researchers.

By the tenure process selection bias I mean that, not only does tenure encourage people to do less risky, shorter-term, easily-assessed research before they get tenure, but also the selection of which people get tenure favors choosing people who inherently prefer to do this sort of research and who will continue to do that sort of research after tenure. That is, if you have a population of 100 researchers, 50 of whom like short-term, low-risk research and 50 of whom like long-term, high-risk research, and only 70 of them are chosen for tenure, the pool of tenure-acheivers will be predominantly composed of short-term, low-risk people.

So, it's not just that those who prefer long-term, risky research have to delay their aspirations until after tenure; in addition, less of them will get tenured at all. The field loses out not only from the years that these people spend not doing the long-term, risky research that they want to, but also from a distortion in the preferences of tenured faculty.

Why would giving tenure earlier change this? There would still be a selection bias, and there would still be a selection. However, at the time the tenure decision was made, there would be less information to go on. The decision would hence be noisier. Therefore, the biases of the tenure selection process would have less impact on the composition of the resulting pool of tenured faculty.

More fun may not be seen as a benefit worth discussing, particularly if you think the happiness of researchers is of little value compared to the value of the results of their research, but I felt it should be mentioned. Clearly[[image:unless of course you believe that the fun of getting to watch younger researchers suffer after you achieve tenure is even more fun than getting tenure earlier yourself would be -- I doubt anyone actually thinks that though, although it's a funny idea.?]], getting tenure earlier rather than later will tend to make life more enjoyable for people. I won't rely on this in making my argument, however.

Rebuttals to the advantages (and rebuttals to those rebuttals)

Even if one accepts that earlier tenure would lead to more long-term, risky, and speculative research, there are some who would say that this would be counterproductive.

Long-term, risky, research is harder to evaluate, and some may argue that it is more productive to focus on more tangible results; that by building small, tangible gains we will waste less effort and so ultimately move quicker than by trying to jump ahead with more dubious ventures.

Some would say that, despite impressive-sounding examples of iconoclastic research that turned out to be right, on the whole time and money would be better spent if individual researchers did mostly what the rest of the academic community thought was valuable, and avoided research that the majority of the community thought was a waste of time{{But not if you put it like that --- people won't consciously identify themselves with the arrogant-sounded cynics that you find in stories of past scientific iconoclasts. To assess a person's feeling about iconoclastic research, one could instead ask them, "In general, is it a bad thing for a grant to be given to someone for work that most people think is irrelevant or wrong, even if the work is not totally ridiculous? Or are you happy that at least some money is allocated to this sort of thing, even at the expense of other projects?"}}.

Speculative research is seen by some as flaky and a waste of time.

The same people who believe that we already have too much long-term/risky/iconoclastic/speculative research will also believe that we have too many researchers who approve of that sort of thing, and so will see no problem with the tenure selection bias against that sort.

A detailed argument in favor of long-term, risky, iconoclastic, and speculative research is outside the scope of this essay; although I have provided some evidence above that others think there needs to be more of this stuff.

Even if having researchers who are passionate about their work is good for the field, it is possible that the research independence provided by tenure is not necessary for passion. Perhaps after a month or two of working on it, a person becomes as passionate about their second-choice, "safe" research topic as they would have been about some favorite topic that they don't dare to pursue until after tenure.

As for empowering younger researchers, it is possible that the competitive nature of the system is the only reason that younger researchers give more time to their work. If there is nothing else special about the younger years of work, then making tenure happen earlier would not reap any benefits by giving more independence during those younger years.

Data supporting the contention that people work longer when they are younger includes this census chart (which is not specific to academia). However, this doesn't really prove the point because one would expect that there is more pressure on younger workers in any career. NSF infobrief 06-302 provides data that seems to argue that the presence of absence of tenure is the only determinant of hours worked, not age. However, I speculate that if tenure were given earlier, that data might look different. First, just because younger people work harder with tenure is no reason to think they wouldn't also work harder without it. Second, it is no good to compare ages within the non-tenure track people and to extrapolate to what would happen with the tenure-track people if tenure were given earlier, because the non-tenure track pool may be less ambitious to begin with. Third, it is possible that the hard work done during the pre-tenure years "burns out" some people to some extent, and that they would actually work more over the entire course of their life if they had not been compelled to work so much for one part of it.

However, while these are reasons why my contention about younger researchers might not be wrong, they are hardly enough to convince a skeptic that it is actually right.

Disadvantages of earlier tenure

The first disadvantage of early tenure that probably pops into one's head is, without a long and discouraging "tenure gauntlet", how do we screen out people who aren't willing to work hard?

The next is that, since not having tenure provides a strong incentive to work hard, getting tenure earlier provides a shorter length of time during which a researcher is pressured to work. Therefore, summed over the course of a lifetime, there is less incentive to work hard. Even people who are hard-working by nature can be expected to work harder or less hard in response to incentives, therefore we can assume that giving tenure earlier will result in some reduction of work.

A second order effect of the "tenure gauntlet" is that, not only does it screen out people who attempt to get tenure but who don't work hard, but that many people who don't want to work hard don't even apply.

Finally, I noted above that by decreasing the amount of time before tenure is awarded, we increase the noise in the selection process and so reduce the impact of the selection criteria on the resulting pool of tenured faculty. This is presented as a benefit above, because the selection criteria are biased towards short-term, low-risk research, but it can also be seen as a disadvantage. Quality, effective, productive research is surely part of the selection criteria too, and the impact of these criteria on the final pool of tenured faculty will diminish also. So, with less information to go on at the time of the tenure decision, it becomes harder for the tenure system to pick out the best researchers.

Rebuttals to the disadvantages

I actually don't think that many so-called{{I don't necessarily think academia should be closed to people who want to work less; after all, many other parts of life such as raising a family are not simply "lazy" but rather are admirable uses of time that should be encouraged; and I think including people with diverse interests broadens the perspective of the field. However, I am sure there will be some disagreement there so I don't want to rely on these arguments.}} "lazy people" are "filtered out" in between the beginning of graduate school and the awarding of tenure. It has been my experience that most people who even begin an academic career love their field and are willing to work at it[[image:Although I have not yet seen this firsthand, I predict that what determines who stays and who leaves later depends less on work ethic and more who is interested in business, who has a spouse who's job is compatible, who gets lucky landing a good position, etc.?]]. So, I think the proportion of "hard workers" in the pool of tenured professors is similar to the proportion of "future hard workers" in the pool of graduate students.

I agree that earlier tenure must, summed over the course of a lifetimes, provide less incentive to work hard, and that this will necessarily tend towards less work. I feel that this reduction will be worth it.

I agree that the earlier tenure will also attract a few slightly "lazier" people to academic careers. Again, I think this effect will be compensated by the other advantages.

While it may seem that having less information at the time of the tenure decision necessarily leads to a lower-quality decision, in fact this is only necessarily true when the decision criteria exactly match what you really want. If what you really want is high-quality, creative, hard-working researchers, but your criteria can only measure that indirectly through publications, then it is in fact possible to reach a point where reducing the noise in your decision process actually worsens the results.

Imagine that there are three types of researchers; "bad" researchers, "good, low-risk" researchers, and "good, high-risk" researchers. Say that ideally, you want to select a population composed of 50% "good, low-risk" researchers and 50% "good, high-risk" researchers. Associate with each researcher a quantity Q. The Q that any indiviual researcher has depends on which type they are; 1 for "bad" researchers, 10 for "good, low-risk" researchers and 9 for "good, high-risk" researchers.

Now let's say that in order to decide tenure for a given researcher, you take N samples from a Gaussian distribution whose mean is Q, and use those samples to form an estimate Q' of their Q. In order to distinguish between bad and good researchers, the chance of a given researcher getting tenture increases as their Q' gets larger [[image:you might protest that you want to use a different method of deciding who to hire. If instead if you pick an appropriate number between 9 and 10 and hire people who fall sufficiently close to that number, you can have your cake and eat it too. But in the metaphor, this would correspond to something like giving tenure to people who publish a lot, but not to people who publish ''too much'', because low-risk people publish more than high-risk, and you're trying to counteract the bias towards low-risk people. But I don't think anyone would actually use that sort of hiring strategy, and anyways, you'd need to have a lot of detail on the exact relationships of your measurable quantities to your true desired quantity in order to pick the right cutoffs, and this level of precision isn't available for the fuzzy criteria used in the real world?]] [[image:I think I am also making some assumption about the specific form of the way that the probability of hiring depends on Q', but I haven't worked this out yet for the first draft?]].

This is a metaphor for the way that the tenure process cannot measure exactly what you want it to. You want to treat low-risk and high-risk researchers equally, but because you cannot measure what you really want to ("high quality-creative-hardworking-ness"), the measurement you end up with (Q) can have some bias towards one or the other.

The amount of time before you make the tenure decision corresponds to N. As N gets larger, you have more information on a person's performance, and your Q' estimate gets more reliably close to that person's true Q value. But past a certain point, that increased precision has little effect on your capacity to discriminate between "good" and "bad" researchers, but a lot of effect on your bias towards "good, low-risk" researchers at the expense of "good, high-risk" ones.

Therefore, if you assign some utility to weeding out "bad" researchers, but you also put some utility on having a fairly even mixture between low-risk and high-risk researchers, then there is some optimal value for N. As N gets larger than that optimum, you will in fact be losing utility, because you'll be killing off the "good, high-risk" researchers without weeding out many more "bad" researchers.

I conjecture that it's actually pretty hard to distinguish between good and bad researchers at all; only near the end of their lifetimes has enough time passed to begin to assess the impact of their work. In such a situation, the impact of the "biases" like the riskiness, short-termness, etc of their line of research on measurable quantities like volume of publications and citiation rank is comparable to the impact of the desired quality on those measurable quantities. Therefore, you would expect to hit the optimum when N is very low (basically, the optimal policy would be to give tenure to every Joe who can manange to turn out even a few interesting publications; or, if you don't have enough spots for that, to decide randomly between them; the rationale being that if you collect more information, you will lose more in diversity of research styles than you gain in quality).

Conclusion

Giving tenure earlier could substantially increase the amount of research which is risky, long-term, integrative, or cross-disciplinary. It would also increase speculative and iconoclastic research. It would also give people more time to work on what they are passionate about, which may produce higher quality work.

I assert that these effects would have a huge positive impact on academia.

It is likely that that the total amount of hours worked would slightly decrease due to the reduced number of years that tenure-track people are "under the gun", and due also to the additional attraction of the career to people who don't want to work so hard. The work decrease may be somewhat but not completely ameliorated by the fact that people would be able to start working earlier on the topics they are most passionate about, and the likelihood that people work harder on topics they are most passionate about.

I believe that the value of more risky, long-term, integrative, cross-disciplinary, speculative and iconoclastic research would outweight the slight decrease in total number of hours worked and that the rate of progress in academia would significantly increase were tenure given earlier.

Postscript

It is worth noting that one reason it takes so long to get tenure is not only that people think that "it should be that way", which is what this essay tries to change, but also that there is simply an abundance of well-trained people without tenure competing for tenure spots. If the community was convinced that earlier tenure was a good idea, the way to achieve it would be to reallocate university finances to create more faculty chairs at the expense of less postdocs and graduate students (less postdocs would be needed anyway, because more of those same people would become faculty).