#197 The Evolving Role Of Human Expertise In Data-Driven Fields, with Melanie Johnston-Hollitt, the Director of the Curtin Institution For Computation

On today’s podcast we have special guest, Melanie Johnston-Hollitt, the Director of the Curtin Institution For Computation, to discuss with us some of the bleeding edge ways that data is being leveraged in the academic space.

For example, a new radio telescope being built in Australia and South Africa will give us new insights into the cosmos. It will also generate 160 terabytes of data per second; an eye-watering amount of data that poses unique challenges about how it’s utilised and managed. As Johnston-Hollitt mentions, where most wisdom says to store all the data collected, in this case, the research teams are better off developing ways to process the data as quickly as possible, and then removing it to make a fresh set of observations.

This understanding of how to best manage and interpret data highlights the ongoing role that data specialists play at a time where automation is taking ever-more amounts of mundane work off the hands of people, Johnston-Hollitt adds. Data automation will achieve three things in workplaces, she says:

1) It will take the drudgery away from roles, allowing professionals to focus on higher-level and more engaging & rewarding work.

2) It will supplement and complement, but not erase, the expertise of humans. Johnston-Hollitt points to how data can be used to support medical diagnosis for less common conditions that a doctor might not see frequently, but ultimately, it’s up to the doctor to make the diagnosis.

3) Data and AI will also result in the creation of new jobs, as people develop more sophisticated algorithms and need people to validate the applications and results.

Hear more detail about all these insights, and more, by tuning in now. Thank you very much to Johnston-Hollitt for guesting on this podcast.

Thank you to you our sponsor, Talent Insights Group!

Join us for one of our upcoming events: https://www.datafuturology.com/events

Join our Slack Community: https://join.slack.com/t/datafuturologycircle/shared_invite/zt-z19cq4eq-ET6O49o2uySgvQWjM6a5ng

“Productivity is about the willingness to do the work and being supplied with the tools that you need. If you're going to be lazy in your workplace. It doesn't matter if you're on the job or remote, you're still an unproductive worker. It's not about location. And it was only that COVID came and forced that change. External force came in and remove that human barrier, that human fear that allowed us to actually remote workers remotely as we are now, even though we've had that technology for 20 years. You see these system levels and also individual levels. And it's a really interesting question to sort of being able to, how do you convince at the system level, a whole bunch of people to do what you're doing, and review their own biases and heuristics?”
- Melanie Johnston-Hollitt, Director of the Curtin Institution For Computation

WHAT WE DISCUSSED

0:00 Introduction
05:45 Change management, the aligning of incentives, the convincing and getting momentum are often the biggest barrier in technology, adoption and progress of innovation in organizations. Is that something that you see a lot as well?
13:39 What are some ways that you started to use on a recurring basis to identify, and then start to provide some structure?
24:41 How do you develop consensus from different opinions?
28:40 What type of projects have you guys been working on? Or deploying? When there’s slower than expected adoption in some of those how do your three areas apply?

EPISODE HIGHLIGHTS

• I remember hearing about the number of petabytes that they're getting from telescopes and having to sort of whittle that data down, to find ways to then be able to manage it. I think there is there's so much overlap between the two fields, and it's really exciting.

• In radio astronomy, we produce ridiculous amounts of data. So the next generation radio telescope, which is called the Square Kilometre Array, which is just to start construction at the moment in Western Australia and South Africa, is going to produce in combined 160 terabytes of data per second raw data.

• I'm really interested in how you extract knowledge out of different datasets for different purposes, and the barriers to doing that, which are often not technological, but actually, human factors. The human barriers. That's what I'm interested in.

• What I find is people internalize messages. And they don't really realize why they're internalizing them. And then 20 years down their career, that becomes a barrier to technology.

• If the objects that you're observing change or the timescale of millions of years, you're actually better off, taking the data, processing it as quickly as possible, throwing it away, and then re-observing. So effectively, you're using the sky as your archive, and as long as the timescale of the objects you're observing is longer than the archive, reprocessing time, so how are you gonna go back and re-observe, that's fine. But it's hard to get people to think about it that way. Because they've internalized this view that data are precious. So it's really interesting to run into those types of situations.

• There's that barrier of adoption because we're accustomed to doing things the way that it's been done, even though there are new generations coming in that haven't seen what was there when that layout was designed. At one point, AI will be the way that we interface with computers, and then keyboards will be a thing of the past. But I think until then, we're stuck with the layer that we have.

• I do two things. So the first thing is, I usually ask people why. In the case of my peers who've internalized this view that data are precious, and do not realize that that's an economic argument that was valid 20 years ago, and isn't valid now. Potentially. I asked them why they think it was the case. And I try and get them to sort of step back and look at the entire system. And, to realize that that was an economic thing. And to think about whether or not they can overcome that heuristic. The other thing that I try and do with people is I talk about what their job is holistic, how AI and machine learning can take away the drudgery parts of it, and that their job isn't going to go away. But it's going to evolve and evolve for the better.

• I'm just gonna say I think there are three things that I will hopefully happen. So one, some of the drudgery will go away from jobs, but people will be able to do that higher-level stuff, too. You'll be able to augment people's ability to do their job. 

• I think we're going to see the rise as I said of data validators, some of them are going to be specialist stated validators for a particular industry. And some of them are just going to be fit particularly for image-based stuff.

• I think anytime you put humans together, we probably have difficulty getting 100% consensus. And then you really have to look at what your business outcomes are, or your research outcomes, and then pick some rules and heuristics based on where you want to get to. But that I think, it’s also an interesting field. How do you get humans to agree to things? That's really difficult. And so ethics around AI and setting up heuristics for AI ethics, I think is a really, really challenging field.

• The other thing associated with that is that the legal system is slow to move. And so we are in the regime where the laws that we have, and legislation is often falling behind what the technology is able to do. And so there's a real impetus, I think, for legislators to have a better understanding of what technology can do or what technology is already out there. 

• From my perspective, now, it's really important to touch base with the human side of these things. So even if people are coming from the technology perspective, you need to have an understanding of the humans that are going to be using the technology and are going to be benefiting from the technology, because those two things are what create these barriers. And so what I would like to see is when data scientists start to work on processes, with clients, and so forth, we don't just ask the technical questions, we don't just ask, what's the scope of the project? What are you trying to achieve, etc. from a process point of view, we ask the questions, who's going to use this? What can we talk to them about? What are their experiences? What are their limitations? Because I don't think there's any point in spending effort on creating new solutions and new technologies, which are then not uptaken. 

• I think if we can talk to executives more so that they know this stuff, any organization that wants to let technologists and data scientists talk to their executives, big thumbs up, please do that. Good for you. It's good for us. helps the world change.

---

At Data Futurology, we are always working to bring you use cases, new approaches and everything related to the most relevant topics in data science to help you get the most value out of these technologies! Check out our upcoming events for more amazing content.