Does Data Influence The Way People Lead?

drawn image of calendar icon

May 23, 2024

hand drawn writing icon

Anastassia Lauterbach

Interview With Anthony Scriffignano, Chief Data Scientist, undisclosed Aerospace Company, distinguished Fellow Stimson Center, Board Member, US Chief Data Scientist of the Year 2018.

Lauterbach: Is leadership a skill set, a science, or an art?

Scriffignano: As with anything, you can study and read about leadership, but until you practice it, it doesn’t become uniquely your own. The most effective leaders are reflective and capable of pausing and thinking about what they do and why they do it. Leadership is a bit of an art, too. The way people lead in emergencies is often very different from how they would perform if they had plenty of time. Being capable of adapting is vital here.

Lauterbach: Does data influence the way people lead?

Scriffignano: On the surface, there is a lot of positive influence. Information can show us the efficacy of whatever we’re trying to do. Generative AI – and, in particular, Large Language Models – can summarize a lot of material that we might not have time to read and, in certain situations, contribute to better decision-making. Still, data is only a tool, that can either be used to create value or as a weapon. We ought to be careful not to use KPIs and data to build a case against a leader or against the workforce to justify layouts, just as an example. In larger organizations, in particular, people produce the data they need to make the necessary argument. That’s not science; that is confirmation bias. If you go in there with a preconceived notion, you’ll find data that supports what you think is true. That doesn’t support the truthfulness of a picture.

Lauterbach: How do you stay critical to prevent data misuse to confirm biases instead of doing good?

Scriffignano: The first thing is to always start with a question. Learning the scientific process makes us realize how we observe the world around us. You ask an important question, and you refine that question, and then you look at maybe what other people have done to find an answer in similar circumstances. Then, you start to look at your methodology. Only after that do you start to collect data. Such a process takes time. People might find it inconvenient.

Lauterbach: How do you teach people to ask good questions?

Scriffignano: I try to tell stories, for example, about where people needed a better question before they jumped into the answer. Finetuning people’s questions, rather than telling them what I think the question should be, is not always easy. It should be done with humility.

Lauterbach: How do you approach data in a traditional business that doesn’t have experience with AI technologies?

Scriffignano: The data sphere, the amount of data on Earth, is expanding at an arguably unmeasurable rate. Humans typically do not manage things that grow at that rate very well. So how can you cope? Don’t lead with the data. There are three corpora of data, to put it simply. There’ll be the data that you have in hand and can access to use. There’ll be data you could go and get, but you may have to take time to do so and spend a lot of money. And then there’ll be the data that might exist, but it will be inaccessible, e.g., personal information, like tax records. You put these three types of data in the context of your strategic or operational questions. You investigate how the data you’ve got enables your problem-solving. You challenge yourself while asking whether you work on what is just a convenient sample, or whether you consider inconsistencies and the missingness of the reality. A great example is a survey. There’s a nonresponse bias in people who didn’t provide answers to the survey, so you have to be able to somehow mathematically or epistemologically consider what they might have said and why they didn’t show up in the first place.

Another thing is to be very careful about data talent in your business. There might be a lot of people that have rebranded themselves as data scientists. There’s data in their life but not necessarily science. Assessing multidisciplinary capabilities in a data team is paramount. You require people who understand statistical methods and preconditions of what you try to model for, storytellers who can make the outcome of your analysis relevant to the broader audience. You need people who can do visualization. You rely on people who can understand regulatory constraints of what data you may or may not use. You require people who are, to some extent, diplomatic. They can go and get access to data that maybe your perspective partners or customers don’t want to initially grant or are reluctant to share. In the world of data, skills that have made anyone successful so far aren’t sufficient to carry everyone forward with all the change that’s going on. Reflection and diversity in thinking and applicable skills are increasingly important.

Lauterbach: How do you interview?

Scriffignano: I’m looking for someone I can work with and teach, and for someone who can teach me. I’m searching for a person who can disagree with me and move on after the disagreement is resolved. I would like to have someone to represent you and me when I’m not there. I’m looking for someone who can hear me when I speak and not presume they know what I’m about to say. Those are difficult things to learn in a formal conversation. So, I try to have a more casual dialogue. I want a person to be as authentic as possible in an interview. Easy to say, not so easy to do.

Lauterbach: Do data tools and data technologies change culture in companies?

Scriffignano: We say that we’ll never report to a robot, but the GPS tells us to turn right, and we go there. Outlook tells us to go to a meeting, and we obey. We will take that to the next level, but the “how” depends on the domain. I may not have noticed some recent article in emergency medicine that says the traditional treatment is not practical in this particular case, as an example. So, I learned from a machine. AI, however, is often not so good at explaining itself. In many situations, human experts must elaborate on why they decided differently from what the AI recommended. We might become hypersensitive, on another hand, and constrain our activities to the point where we only do where the algorithms lead us to. That stifles creativity, intuition, and the ability to adapt. Organizational agility calls for a thoughtful and critical approach to how we use AI.

Lauterbach: What is your view on synthetic board directors, those avatars mimicking the decision-making patterns of real people?

Scriffignano: We might have a Metaverse populated with characters such as a risk-averse automaton, a growth addict, or a contrarian. Such a persona sits on almost all boards. In a virtual boardroom, we might assess the comments and contributions of such avatars and compare them to models we try to analyze to make a decision. This might not be a bad scenario. But I don’t want these things to vote. It is leadership that makes us uniquely human. I don’t want to lose this imperfect humanity.

Lauterbach: While admitting my own biases against synthetic data, how do you think about their growth in problem-solving across use cases?

Scriffignano: There’s a concept of elasticity of data that I kind of steal from economics. Consider decision elasticity: how wrong can you be and still make the same decision? And if you can find a way to measure that and then prove to me that by adding your synthetics, we will stay within the same domain of decision elasticity, I’m okay to use synthetical data. But if you can’t or become increasingly uncomfortable, then maybe we should talk before we go full steam into creating additional noise with synthetical datasets.

Lauterbach: What are the significant benefits and pitfalls around the democratization of AI today?

Scriffignano: Democratization of AI will happen whether we like it or not. AI is still like a teenager who doesn’t do exactly what we desire. Sometimes it does unexpected things. How can you mitigate potential risks? You have fully centralized control if you think of a pendulum that swings over one side. And on the other end, you have a fully democratized approach, on another – centralization. I think it’s dangerous to be in either one of those places. If you’re in a fully controlled and centralized environment, AI “super gurus” deliver the answer. In a fully open scenario, anyone can enter the drugstore and get any drug they want without a prescription. The truth must be somewhere in between, and closely connected to the purpose. Technologies and their use will change while you look for this purpose, adjust it, and learn. Laws will also change, along with gathering experience about the unintended use of things. Critical human thinking is paramount to navigating the unknown.
I worry the most about the new, younger workforce. They’ve had access to the tools and perfect datasets that always contain some answers. They might not be critical of the tools and the domains they are working on. We shouldn’t create a race of colleagues who are accidental slaves to the machine because, without the machine, they can’t think independently.

Awareness about limitations of data and tools must be on the menu, whatever we do with data. When we do a web search, we’re searching a tiny fraction of the information available online. Some people believe, however, that they’re searching for all the knowledge of mankind. Does that matter if they look up the capital of a country? It doesn’t. But it matters if they investigate the political context.

Lauterbach: Do we need to rethink how we do AI altogether?

Scriffignano: We should constantly challenge how we think about … everything. If some machine learners believe all they need is training data, and they can predict the future, they’re wrong. They can expect certain features of the future, but not more than that. Chaos theory is in its infancy in terms of implementation. Quantum computers are going to change the way we do things. I don’t know what quantum AI will look like, but it will happen. Change is unstoppable.

The funny thing about change, though, is it’s hard to see when you’re part of it. This is why it is paramount to focus on leadership, which starts with reflection and principles. It doesn’t begin with equations and datasets.

Lauterbach: How do you ensure you can reflect and stay agile?

Scriffignano: I am significantly influenced by martial arts. So, there are techniques in martial arts for calming, centering, and focusing. I don’t use them every day, but when I use them, they help. I surround myself with things that inspire me. Art and real books help. Finally, I like to breach rituals. I don’t drive to the office in the same way every day. It’s probably way less efficient but allows me to challenge myself.

Lauterbach: Is inefficiency a good way to deal with AI?

Scriffignano: Inefficiency – or human inconsistency – is a way to prevent or defend against AI myopia. Playing chess against the computer is very different from how I play chess against a person. If a computer is in front of me, I try to get out of the traditional chess opening book as quickly as possible. Otherwise, it is not a real game.

romy and roby and the secrets of sleep book cover

Book 1

Romy & Roby And the Secrets Of Sleep.


Submit a Comment

Your email address will not be published. Required fields are marked *

Select your currency