Automating the 21st Century Farm
by Andrew Kerr
April 2007
The CEISMC Gazette
In February I sat down with four graduate researchers in Georgia Tech's School of Psychology. We discussed the future of automation (in very broad terms, turning certain tasks over to robots/computers and putting the humans in more of a supervisory role over those robots/computers). We began by focusing on agriculture (panelist Neta Ezer had interned at John Deere), and then wandered into other territory.
Participants were Marita O'Brien, Neta Ezer, Andrew Mayer, Bart Wilkinson, and myself.
Neta - A lot of people think of John Deere as "the tractor company." But John Deere devotes a lot of time and research developing technology. They're doing a lot of robotics. Where we come in is in looking at what's going to make robotic vehicles easy to use. If you automate a tractor to be able to drive in a straight line, then a person who doesn't have as much experience can still do that and can focus on other critical tasks at the same time.
Gazette - When most people picture a stereotype of a farmer, they probably don't picture a computer wiz. How readily and enthusiastically are farmers embracing technology?
Neta - One of the reasons that people go to John Deere is because they heard about this new technology that is on the vehicles that other companies don't have.
Marita - For the farmers who are successful, they're running a small business out there, they're getting all their information about their crops from computer systems. The successful farmers are really having to invest in this technology.
One of the challenges is that some of them are older. The average age of a farmer in the U.S. is about 55, and so it's part of a generation that may not have grown up with computers. One of the reasons that we work with these issues for John Deere is because we do a lot of research with older adults.
We found that there are a lot of older adults who use email, who will use computing systems if there's a benefit for them, like communicating with their grandchildren, or being able to see what their grandchildren are doing in school. But teaching them how to use them is different.
Bart - I come from a part of the country that's very agriculturally based--Nebraska--and I can say from a little bit of experience working with and being friends with people that were in the business or who owned family farms that once a piece of technology caught on with somebody it spread like wildfire, because nothing affects their bottom line more than being more productive and more efficient. [Running a] family farm would be, to me, the most stressful job in the world, because every winter you're mortgaging all your profits from the last harvest to invest in equipment and herbicides and seeds, and you're putting it all on the line every growing season, every harvest, just kinda hoping nothing bad happens or that your predictions are right and that your investment in technology is right. So once they find something that works predictably then it spreads pretty quickly.
Andrew - And I think also that the stereotype of farmers not being computer savvy, I think the general population of that age is just not technologically savvy. That points to the importance of good design.
Marita - One of our research areas is technology acceptance--what makes people more likely to accept technology. With the kind of investments that's required for this kind of technology, I mean, tractors, a combine, isn't that almost kind of a $10,000 investment?
Neta - A combine is half a million dollars. [Group laughs.]
Marita - [Laughing] OK, so for that kind of an investment it's gotta work! I can't afford down time. I spent half a million dollars on something that may be productive for only a couple weeks a year.
Andrew - I'm interested in expectations of performance. I think that that's particularly important with John Deere because John Deere does have such a good reputation; their slogan is "Nothing runs like a Deere." So managing those expectations of how it's going to perform is important for people.
Neta - No technology is going to be 100% perfect. So people need to be aware not to overtrust or undertrust it. That's a lot of the research that we have been doing, how people calibrate their expectations to what the technology can actually do.
I'm looking at how people would control multiple unoccupied vehicles. So you have a vehicle that usually has a person in it, but you take that person out, and you could be miles away and be controlling this vehicle. I'm looking at how many vehicles one person can control at a time, and then other variables that are going to be important to the success of that whole system.
Gazette - I'm picturing almost a video game kind of thing, driving tractors remotely. Is that what it's like?
Neta - To some degree it is. The military has done it for a while with unoccupied aerial vehicles, and right now it's two and half people for each airplane. They really want to have one person control multiple airplanes. The technology has been there for years and years, but no one has really pinned down what it would take for one person to control all these planes.
Marita - My research area is more with consumer products. They continue to come out with new versions of cell phones that have an increasing number of features--about 10% of which people will use. How can you allow people who buy things to really make better decisions about what features they want and then, if they wanted to use something more advanced, have the capability for them to figure out how to use that feature without having to go back to the manual which, if the manual was there, probably was not written that well? A lot of people want to sell products as being "intuitive," but there's really not a lot of definition on what "intuitive" means. So my research is looking at trying to get a better model of that.
Gazette - [To Bart] What sort of research are you doing right now?
Bart - I try to add on to the automation research that's been done, continue in that line, what was mentioned before, people's acceptance of technology and expectations of technology. And Neta also talked about cues, and how no automation is perfect, and that there are certain situations that would create errors in the automation. What I'm looking at is giving the user the appropriate representation of those kinds of situations where the automation would fail. What you would ultimately look at is a better way to train and educate users so that they understand when and how the automation would fail.
Andrew - This is more just a general statement, not specific to farming or anything, but the important thing to understand is that adding automated systems just by slapping them into an existing system doesn't necessarily mean that the problem's going to be solved, or that the person no longer has to worry about the task. Usually what happens is that it changes the nature of the user's task. So it might change you from being someone who is constantly looking for cues out in the field to somebody who is supervising a system.
Neta - So still there is a monitoring component of the automation. Is it working this time? Why did it make a mistake? And there's this whole cost benefit ratio. If the system costs this amount of money, but it's not really going to save me any time, then is it really worth using it? Or if it's not going to be reliable enough is it worth using it?
Gazette - I guess there's always the danger of simply exchanging one type of stress for another.
Marita - In automation one of the challenges is to try and understand that the computer may be doing the task differently than a person. So it's not just a matter of here's how the person does the task, I just need to automate that and then the computer can do the task.
The cues that a person uses to try to drive in a straight line may be very different than what a tractor would use. Part of the issue is that people aren't even aware of what some of those cues are that they're using.
Neta - We're trying to develop a model of automation use. We're trying to find out what are the important variables in that model, what are the inputs and outputs. There are lots of gaps in the research, so we're trying to fill in those gaps. Bit by bit we're trying to fill in this model so that we understand the whole process of what it takes for people to use automation appropriately.
Andrew - As Neta said there's lots of variables and there are a lot of gaps, but we can look at them in terms of people variables and automation variables. Self confidence in expertise is a person variable that will affect how much you'll rely on the automation. Then you might look at reliability and type of error that an automated system could make, so you have those all interacting as well. So it's quite a complex picture, and the model of interacting between humans and automation is quite complex because of all the interaction of those variables.
Neta - And what kind of automation? Is it a very simple type of automation where it maybe just helps the person sense some information in the environment, but then the person still has to interpret that information to make a decision, perform an action? Or is the automation fully automating something where the person is just kinda watching it, waiting for the one instance over a long period of time where it might make a mistake? You never really remove the human from that because there's still that monitoring element.
Andrew - And one thing that Bart's looking at is that as automation advances you're not so much looking at automation as just a helpful machine, but you're looking at it as potentially a teammate that you're actually interacting with, and coming to decisions together. So then you have all these variables of team interaction that also interact with these individual characteristics and automation characteristics.
Neta - And increasing automation doesn't necessarily mean that the task is going to be more successful. If you have a lot of automation and people there, just, if you think of just a nuclear planet and it's all automated, and the person's just there sitting day in and day out just waiting for that one mistake, when something happens they may forget how to do things because the automation has been doing it the whole time! Or they may get really bored and fall asleep and they don't notice that something is going on. So we're looking at the balance of level of automation along with person characteristics, what type of person is it, is it someone who is very motivated versus someone who will just let the automation do whatever they want and they'll go and take a nap while it's doing its thing!
Marita - There's even examples just in the in-car navigation systems that you see in rental cars now, turn right here, turn left here, the computer may not have all the information. And what if you say the computer's telling me to do this but I want to go this way, well then how do you tell the computer that and then the computer says well here's your route from here. There's a lot of human factor stuff that goes into saying what kind of cues should the system tell you, how far in advance do you need to know that you need to make a turn, that you need to get over to the right, what kind of information is useful to provide.
Andrew - I'm going a little bit off on an aside, but my dad uses a navigation system in his car, and hated it when he first got it. He'd always complain about it, it told him the wrong things. Well he's been using it now for quite a while. He came to visit in Atlanta and rented a car that didn't have one, and all of a sudden he was getting lost everywhere, he couldn't do anything because all of a sudden this aid that he had become so reliant and dependent on was taken away.
Marita - You know we could probably look at that just in terms of Map Quest. Right now when I want to know how to go somewhere, I just go to MapQuest, this from here to here, here's the directions...
Neta - Yeah I don't know how to read a map at all!
Marita - We could probably study just what the effect is of people who have kind of grown up with MapQuest, versus...
Andrew - So this is kinda sometimes how some of our research ideas come out! [Group laughs.] Just talking anecdotally about things, and we say, "Yeah you know that actually is a problem!" And then what are the psychological variables involved in that?
Neta - And all of a sudden we have five studies at once! [Group laughs.]
Bart - It's interesting what you said about your dad. At first he didn't understand why the automation would make mistakes, but over time he slowly gathered this mental representation of how the system worked and he began to understand its needs and outputs, and then finally he got to a point where he trusted it, relied on it. Wouldn't it be great to be able to give him all that information before he even started to use it? So that he just went from not using it to understanding, having a mental representation of its workings, its inputs and outputs, and wouldn't have to suffer for, what? A year?
Marita - I think there may be some person variables in there.
Andrew - Oh yeah. There's definitely some person characteristics involved in that too. But it's interesting that he obviously had to change his mental model to adapt to the automation, but when it was taken away there was a long curve for him to be able to go back to his pre-existing mental role.
Gazette - Seems to suggest that these systems should, while making everything easier on the one hand, be able to provide some level of useful instruction on the other.
Bart - The question designers need to ask is how do people mentally represent their environment, and then how do we design the product so that it matches the person's representation, as opposed to forcing a person to change to the automation?
Neta - People who are making the systems, they're not the users of the system--they know the system way too well, they know how it functions, they know why certain buttons are placed each way. It's when you bring in people that are going to be using the system that you realize that it's not as intuitive as the engineers and designers thought it would be.
Gazette - How did you each wind up at Tech's school of psychology?
Neta - I was an undergraduate here at Georgia Tech studying industrial design, and I was very interested in psychology, so I took a lot of psychology classes. I saw an opportunity to be an undergrad assistant in this lab and I just learned about the different research. I worked with the first person in the lab who was doing automation research--he's actually an employee now at John Deere. I just got very interested in it and applied and here I am, four years later, still doing what I like to do!
Andrew - I'm originally from Canada, so I was doing my psychology degree there. I took a class in human factors. I started working in the prof's lab back in Calgary. I knew that I wanted to go into this area of engineering psychology--human factors, and so just kind of looking at different programs and what their interests are, and people ,and what kinds of publications and research they're publishing, that's how I chose Georgia Tech.
Bart - I'm an active duty army officer, and I'm here on a funded graduate school program, but I actually did this major as an undergrad at the academy.
I kinda knew that a lot of people think of psychology as a very, I don't want to say a "soft science," but they don't associate it more with a lot of the "hard" sciences, like the chemistry and biology and stuff like that. This was especially one of the reasons why I wanted to come here. I knew it [Tech] was an engineering school and it was a very hard science type of place, and that this department is also in line with that. You have to be sound in your psychological principles, but you also have to be able to analyze the data, and in analyzing the data you have to be sound in statistical principles and more hard science type stuff. And I think you'll see, these guys can speak more to it because they're read way more research articles but a lot of the senior faculty here are very sound in stuff outside of just you general psychology core, like quantitative methods and stuff like that. So that's what kinda drew me here. A lot of people have misconceptions about psychology.
Gazette - It's very different from having a person lie on a couch while saying, "Tell me about your mother."
Neta - [Laughs.] That's what most people think of psychology! When I tell them that I'm studying psychology they're like, "What's wrong with me?" And you have to explain that, yes, there is that aspect of clinical psychology, but there's also lots of other types of psychology. We don't have clinical psychology here at Tech at all. We have people who study psychology at organizations and jobs, we have people that study engineering psychology, how people interact with machines, and we have experimental psychology, cognitive neuroscience, how the brain works...There's so much more to psychology than just having someone lay on a coach and tell you about their mother.
Andrew - The classic joke I hear when I say I'm studying engineering psychology is that they think we put engineers on the couch to figure out what's wrong with them! [Group laughs.]
Marita - One of our advisors is a kind of a forensic psychologist, so he looks at, if there's been an accident, were there some factors in the environment that contributed to the accident. Were the warning signs not placed at the correct location, if there was a construction zone had they not set it up so that somebody so that somebody knew they had to get over early enough that they avoided the construction workers standing there, what are the kinds of things that contribute to that? So he helps where there have been some legal trials to try and identify what the problems have been. Some people work for the government to set up the standards for sign placement, sign design, warning design, and warning labels.
Gazette - Is there any end in sight for the kind of research we can do? Will there be a point where we know the human mind perfectly?
Neta - Well compared to the other hard sciences, psychology started a lot later, it's only been around for little over a hundred years, compared to chemistry that's been around a couple hundred. And we still don't know a lot, especially about the brain, they're still working a lot on that. We don't understand how the mind works, what's consciousness, things like that that are so basic to us. There's still so much to do.
Bart - One of my professors that I'm taking a class with right now just about every week says there's still a lot of unaccounted for variance in psychology. And he's been doing it for a long time.
Gazette - Are we becoming smarter in general? It seems that with all the technology there might be changes in our intelligence. Or does the level of intelligence remain the same, just different?
[Long pause.]
Bart - That's a really good question! [Group laughs]
Neta - I wrote a paper for a class last semester where apparently there's something called the Flynn effect, where people's scores in IQ tests have been rising steadily for the past fifty years. But they don't know if it's just an artifact of more people getting an equal education. It doesn't necessarily mean that people are smarter; maybe they're just better at taking tests!
Marita - And the cultural issues, we're still in a very technologically advanced society, so there's so many more people who don't have the technology we have.
Neta - There are very few jobs now that don't use computers in some form, so you really need to be adaptable to changing technology. So even the farmer now has to use all this computerized equipment.
Gazette - I sometimes wonder, What did my dad do 50 years ago at his job? Did he have a typewriter? How did he get things done?
Bart - I think people used the phone a lot more, that was one of the big things.
From the military perspective, in the past five years I can't count how many times people lost their heads because coordination was done over email. There's just no certainty that somebody reads an email. There is if they do confirmation send or whatever. But when people start relying on technology, it just takes responsibility off them, and leaves it out there, and puts responsibility on another person, responsibility that the other person doesn't even know they now have. Before, you sat at your desk, phone rings, there was direct coordination even if you weren't face to face.
So, it fixes some problems but it creates some. So what's better, talking face to face, or voice to voice, or to send off a quick email? There are various problems associated with all new technologies.
...which should keep Georgia Tech's School of Psychology busy for a long while!
© 2008 Georgia Institute of Technology :: Atlanta, Georgia 30332