AI Ethics in Construction: Building Responsibly

Speaker A: Foreign welcome to the Everyday PM podcast. The podcast where we discuss project management principles for your everyday life. My name is Anne Campia. I am the host and founder of the Everyday pm. And today we're diving into one of the most transformative and could potentially be a little bit controversial topics in construction project management. It's the intersection of AI, artificial intelligence ethics and the future of planning and scheduling. Our guest today is Dr. Mala who is a senior PM professional with over 17 years of global experience leading planning, scheduling and project controls for complex infrastructure programs exceeding $47 billion. Dr. Mala is joining us for a series of these podcast episodes where we are going to really dive into various topics on construction project management. If you have not been introduced to Dr. Mala in a previous episode that you've listened, Dr. Mala, why don't you take a second to introduce yourself to our audience?

Speaker B: Yeah. Thank you so much Ann, uh, for this wonderful topic that we are going to have conversation with and um, giving a very brief overview about myself. I've been in the construction landscape for over more than 17 years as Ann has mentioned. And with respect to the kind of diverse projects experience, especially in the project controls roles, starting from the scheduling, planning and the various sorts of contracts management operations, I bring in the wealth of different perspective lenses and along with this the tools with which um, the importance of the various control systems that are necessary for projects, especially the mega scale and mega infrastructure projects, is something like which I have gained over a period of time. And the tools, especially with respect to technological advancements which is going at a meteoric pace, is something which we need to engage not just ourselves but also the peers and colleagues in different levels of maturity with which they want to have to get in the AI related ream to their specific domains. So uh, apart from this I would like to also add that I'm not just only focused on the practical aspects of the industry related experience, I also have gained the research acumen through my doctoral studies. So it's a combination of scientific way of analysing, analyzing the various types of projects that I have accomplished. Yeah, uh, and uh, it's, it's a combination of research theory plus practical implication is something which I'm honored to have got over these 17 years. So without much ado, if you want to know my profile a little bit of more extensively, I hope you would be in a position to replay the episode one which was discussed on construction workforce necessity in the US industry and US uh, USA construction industry, specifically the mega projects. So I have explained um, a bit more about It. So in today's, um, episode, we will have some sort of conversations with the kind of, uh, AI exposure that I have put into the practical applications and how does it, uh, what works and what didn't work for me is something like. Which I would like to, uh, discuss about.

Speaker A: Yeah, absolutely. And you know, the thought of AI, especially in construction project management, you know, as we start to integrate AI more and more into our project workflows and we see the predictive analytics that come from AI, there's questions that project managers are facing now, right? Like how do we ensure the tools are used responsibly? Who's accountable if AI makes a mistake? Right. We can't just assume AI is correct every time. How do we balance efficiency with ethical considerations? I think all of those are things that project managers are curious about, especially in the realm of construction project management. So I'm so excited, Dr. Mala, that you're here to bring that expertise as well as the academic side of it, plus the practicality of using AI, uh, with project workflows. So why don't we dive into what I wanted to ask you first. So you had mentioned previous in the first episode of this series as well as today, that you've worked on some very, very large construction projects, very complex, uh, large budgets. So you've published research on AI applications in the AEC industry as well. So let's start with a quick reality check, Dr. Malla. Where is AI actually making a difference in construction planning and scheduling right now versus where it's still just this kind of hype or this myth around AI? And can you share a specific example from your work where AI has genuinely improved project outcomes?

Speaker B: Well, that's a really, uh, very great place to begin with because the very moment when a person mentions of AI, um, at least for five years ago, I mean, like pre Covid, at least in the construction industry, it was a kind of a, um, um, buzz. It wasn't that kind of a buzzword or it was a mere thought like AI couldn't penetrate or get, uh, get assimilated into the construction industry. Oh, it's just the IT folks. It's not the construction industry professionals who, who needs to learn. That was the kind of notion pre Covid. It was there and it so happened at a skyrocketing pace. The hype and the kind of implementation with which AI is happening in different domains, not just in planning and scheduling, but in the different, uh, life cycles of the construction projects. Right? Starting from the conceptual stage. Initiation, planning, execution, monitoring and controlling, closing and Even maintenance you have the kind of use cases built to this in, in, in just a span of five years. I mean at least there is a surge of interest that has been shown vehemently by various construction professionals. And coming to the uh purview of planning and scheduling systems I would bring like from my experience both as a practitioner as well as a researcher I'd say that uh AI is not revolutionizing the construction per se but it is meaningfully improving the various parts of the planning components, coordination and decision support where the fundamentals are absolutely intact. And it has got good data. So having data is not the right way with which AI can be implemented. But having the good data, data points, data structuring, information structuring is something which is key to implement any sort of um, AI related models or something like that. So if I look at where the real value is uh being delivered in the various projects that um have been recently involved it's primarily in the areas of uh, building information modeling based virtual prototyping at the early project stages. So this is more of like um four dimensional bim. So a combination ah of the project schedule with the three dimensional model and trying to integrate the various information components and trying to look at the sequencing and at the conceptual level how it's going to get built is something like, can be visualized. So that's the first point. The second one is like key is the information structuring and ensuring proper interface management is uh established in the projects. And the second uh, the second most important aspect, construction industry, especially the various projects, every day they produce humongous amounts of data in the form of daily project report, lots of information there and lots of progress site photographs. On this sort of information that's being generated in the project site at daily basis is something where this information, whether it is cleaned properly so who is going to verify the reports. I mean like up until now majority of the construction projects were like maintaining this sort of information or database, more of like auditing, more of like mere compliance kind of system. Yeah, so there was no sort of compliance of the data quality that happened. But with the uh, kind of surge in AI related um approaches that we want to implement it's necessary that data quality is of immense importance. So that forms the basis with which the various algorithms that we can apply to the data that has been generated from the projects. So the third most important component which um, I can think uh, it has delivered uh with respect to AI was identifying the patterns in the schedule activities with respect to risks and any sorts of trends in the historical project data. Although this can be done performed even on Excel, Microsoft Excel and other tools that are commercial tools that are available with the help of AI, different types of algorithms using different um, uh deep learning techniques or natural language processing uh techniques and all we can perform lots of, we can dissect the data and visualize it in different manner so that the visualization and scenario testing is something which can be performed aptly rather than just going with autonomous decision making. And this is where the fourth point that I would like to highlight how it has um, impacted and got better deliverables. So the first one is BIM based uh virtual prototyping in the initial stages of the project. Second one is the information restructuring and managing of the interfaces. Third one is it's able to give us various sorts of anomalies through the pattern recognitions. And the fourth and a final one is like the visualization and scenario based testing. Yeah that if we have humongous data. I mean like it's. It's quite interesting that all these years majority of the construction projects were maintaining databases which weren't of good quality. I mean data cleaning and all was not that um, of utmost importance or sort of um, accuracy needs to be maintained. Maintained is, is not that kind of a notion that was prevailing. But if you want to implement AI systems and it's necessary that data needs to be of good quality and that's where the key point is. So with respect to my research arena I would like to discuss whether it's uh, a agile based BIM or the building information modeling or whether it is lean agile integration which I have performed uh research it had shown significant improvements in the outcomes which is not AI related stuff because that happened purely owing to the data discipline integrating the process based systems and more clarity of the information flows. So for instance I can give an example like one of the water infrastructure projects that I worked six years ago, uh the BIM models that were used for virtual prototypes combined with the structural um, various sorts of um, um schedules or the planning logic had significantly reduced um downstream level rework or the coordination failures because we are in a position to visualize it prior to its breaking down on the ground. So there wasn't a sort of neural network or AI kind of stuff being involved. However this uh, virtual prototyping um definitely had helped in decision making. There was intelligence that has been embedded in how the information was created, validated and shared. And uh, whereas how I see the hype where AI is positioned is as a substitute for a project understanding. So Bringing in AI is not trying to bring in a human. So you cannot substitute humans judgment.

Speaker A: Right.

Speaker B: So no algorithm is going to compensate a human judgment compensate for poorly defined scope or weak contract management or immature project governance. So AI can be considered as a force multiplier or an amplifier to your good systems, but it cannot definitely take away the jobs. So AI is already valuable as a decision supporting tool over here, but only it sits on top of strong processes that are governing which is the information management especially you have the various sorts of BIM standards for information management. Uh then there are also lean thinking principles and human judgment definitely adds um, uh, key role while implementing AI systems. So this is what my take on Ann.

Speaker A: Yeah, yeah, absolutely. I think in a nutshell what we're learning is it's AI is available, it's there, but it's not completely there to take over our jobs. I kind of think we mentioned that even in episode one as we were talking about um, exactly that topic as well. So I'm curious as we dive a little bit deeper because we need to address the elephant in the room which is who's responsible when things go wrong. Especially as you're working with AI systems to let's say do project planning, project scheduling, making decisions, analyzing the delays, um, predicting project risks. Decisions that can affect billions of dollars and thousands of jobs. Right. Especially on some of the projects that you've uh, had the opportunity to work on. So I know you've done extensive work in forensic schedule analysis and delay claims. So how does AI change the accountability landscape? And what are some of the ethical frameworks project managers should be considering or thinking about?

Speaker B: Yeah, yeah, this is really a very um, close. This is a kind of um, an interesting and uh, debatable question also.

Speaker A: Yeah.

Speaker B: So at one, on one hand um, majority of the organizations are um, craving for embracing AI into their systems. So if you don't implement AI it seems like many of their counterparts would be um, of the belief that we are outdated.

Speaker A: Yeah, that's definitely the sentiment we're all feeling. Right? Yeah.

Speaker B: On the other hand there's another notion that whoever is, I mean that's trying to promote or embed AI is also at the cusp of the integrity of the data, project data then non disclosure clauses, NDAs that has been signed by the various project professionals in the construction industry. And what sort of um, um ethical validity are we working with? Like what sort of ethics with which I'll uh, be working with when we use these AI systems? Definitely accountability is something like, which we give to a Person. Right. So when you try to utilize AI, we cannot put a blame stating that this decision has been given by AI. So AI needs to be accountable. Okay, then to which penal code do we have to onto that person? Like it. It's so weird to uh, get to this point. However, a lot of us have um, this sort of notion. So I would like to throw the kind of um, the kind of experience and exposure that I have while I'm utilizing AI. So this is definitely a very um, ah, question which is very close to my um, domain of experience, especially the forensics delay analysis which is completely dealing with um, claims disputes, litigation and various sorts of um, uh contractual obligations. So in this construction industry, decisions are not just simply um, quite ad hoc or abstract. Any decision that's taken is affecting the money, safety, various sorts of um, reputation of the organization and the, and livelihoods of the various um employees who are living on, on this industry. So accountability cannot be outsourced or been thrown to an algorithm which is developed by an AI or something like that. One of the dangers I see with the growing tendency of how AI systems is being um, uh, AI outputs, AI systems being utilized is uh, the objective truth. You give certain information context to an AI system or a large language model and it brings out to you deliverables or responses. So it's, it's a person's judgment to analyze the output with which the LLM has tried to give needs to be validated.

Speaker A: Sure.

Speaker B: It's if, if it is, that depends on the kind of decision that's being made. So I don't feel that depending upon, I mean like correcting minor um, documentation or grammatical or kind of proofreading kind of stuff is something like okay, but then every uh, operation that you perform on AI algorithms or systems, it needs to have a kind of proofreading, a kind of validation.

Speaker A: Sure.

Speaker B: And human judgment cannot just go with AI, AI say whatever it says. But AI systems are only as neutral as the data, um, assumptions or contractual context, whatever that you give or training it on. If the historical data itself is providing some sort of biased practices or adversial contracts or practices or incomplete records, then definitely AI cannot identify whether the data is correct or not.

Speaker A: Right?

Speaker B: The data, I mean whatever you feed and it tries to operate and then give you the output. But how would you know whether the data that you fed is the right one? And if you feel it's the right one also how would you justify the output or response with which it has given that you can rely on? So it's something on the experience, part of your experience, part of your domain knowledge, expertise plays a key role and AI just plays a kind of an assistant to you to subvert some of your uh, time consuming tasks or the tasks which would take level of effort in documentation or providing various sorts of standard procedures or standard operating practices. It's somewhere you can try to utilize certain AI components. And with respect to accountability standpoint, AI doesn't change the fundamental principles. Humans are finally responsible for it. And what does AI change is how responsibility must be structured. AI can help you in structuring your responsibilities uh, between the team. So it gives you multiple options so that you can try to minimize the time spent in organizing the stuff or trying to structure some mechanism. So it's good to have AI to brainstorm. So suppose if you don't have particular uh, team members that you want to brainstorm so you can brainstorm your ideas and probably you may get some sort of leads or some sort of, of different uh, uh, sort of hybrid, hybrid analysis that you wanted to do. Probably that's uh, the what if scenario analysis platform which you can do all sorts of your, whatever that you're thinking, try to utilize it as a platform to do the scenario analysis. And based on my research into this BIM contracts, interface management or the dispute resolution, some of the uh, uh, some of the book reviews as well as one of the papers that I've been involved, a couple of papers, I understand that the organizations need to clearly define who is owning the AI generated insights. So a lot of governance needs to be developed. Otherwise it's like people go to the rabbit hole.

Speaker A: Yeah.

Speaker B: Establishing proper human in the loop decision protocols. Although you're utilizing certain AI tools like for instance in the forensic schedule delay analysis, uh, while we, while we prepare a report, final report on the various sorts of schedule delay analysis that we take into the approach. It's the, the most time consuming part is providing that report. So probably in those scenarios, structuring the report, writing the clear chronological narrative of the events, trying to uh, transcribe the various sorts of narratives with respect to delays uh, given by the site in charge of superintendents, it can bring out the themes, bring out certain causalities, what caused these delays. So you can utilize it as a tool in utilizing in such kind of analysis. And that, that's the part which is consuming a lot of uh, level of effort from the final decision uh, rather than the final decision point. So when the AI systems can be utilized in um, these sorts of minute tasks, it would definitely help the schedule delay experts or for forensic schedule and quantum delay professionals in spending majority of the time in bringing out the analytical component into picture and trying to analyze the substantiation of the cost with explainable and auditable components, uh, with the help of AI outputs that they get. And aligning the AI usage with the contractual and legal frameworks is also another important aspect. So in this forensic schedule delay analysis only suppose there are various contractual entitlement for delays of different types, excusable, non excusable, uh, various types of concurrent pacing delays, all these sorts of contractual obligations that is provided, which is like quite voluminous. It's, it's really helpful when AI systems are used to try to interpret the clause. Not everybody is an um, legal expert, right? Like especially the freshers who wants to join or have a flavor of this sorts of schedule delay analysis. I, I think working with AI and trying to understand the various contractual clauses before it is being entitled to a particular delay is definitely helpful. So when it's simple, like when you have humongous data, AI would definitely help you in different use cases. But identifying that different use case is up to the human to know. And that can come when that particular professional is adept or expertise in their particular domain and has got vast knowledge on the projects that they have worked. Viewing from different perspectives is something like uh, gets developed when you have an exposure on or an experience. Yeah. Ah, so in claims and disputes especially AI should never be the final authority. At best it can help you in bringing out those patterns, themes or inconsistencies. But causality, I mean getting the entitlement, responsibility. But responsibility, who is going to be responsible for particular delay, Whether it is a contractor, owner, sub consultant, consultant is primarily determined by the professionals who understand the context and contracts and the case law. Um, and especially it's not AIs who are working on the project, it's the construction professionals. Right? They're, they're working with the projects so they should be final, accountable people. It's not definitely AI, not AI. And finally I, I would like to add this ethically. Project managers need to ask themselves that. Do not ask just can we use AI? But they should be thinking like should we and under what governance should be working with these AI systems. Yeah. So definitely organizations need to design ethics into their systems right from day one and not just doing a retrofitting kind of um, uh, assessment or just when things go wrong. So before launching AI and sort of understanding the gap, understanding the various sorts of information management standards, processes, workflows and pilot projects are necessary before Launching. That could be a kind of, uh, what if scenario or a kind of scenario testing possibility that could be done. Yeah, Absolutely.

Speaker A: Absolutely. Well, Dr. Mala, that was incredibly insightful. Um, I think you've given us at least the listeners of our podcast, episodes of balanced perspective that couples really well with what we covered in, ah, episode one of this series around building skills from the ground up. I think there continues to be a theme in what you're sharing in your research as well as what you're seeing out on the field, which is the human element and how that can really partner well with AI, but not necessarily acknowledge AI as, you know, another stakeholder in the room that we can hold accountable to if something goes wrong. So, um, I appreciate the balance between perspective, uh, you acknowledging AI's real potential while being clear, uh, eyed about all the ethical challenges that we still need to continue to remember to address as project professionals. So some of the key takeaways from this episode that I would share with the audience start practical, as Dr. Mala pointed out, focus on, uh, some of the AI applications that are working today rather than chasing what the hype of the future of AI is, is going to be. Um, build those accountability frameworks. So before even thinking about implementing AI, you should be thinking about the decision making protocols that Govern instructure that Dr. Mall pointed out. Um, audit your training data and involve diversity in the stakeholders that are looking at the AI system design, uh, and then augment, don't replace. I think for me that was a very strong message in what Dr. Mala shared today, which is focusing on AI as a tool that enhances human judgment rather than replaces it. That was again, incredibly insightful and balanced for what could potentially be a very controversial topic. So Dr. Mala, thank you for bringing that expertise to the podcast today. Uh, for folks that want to continue the conversation with you, where can they find you online?

Speaker B: Well, um, I would appreciate, Ann, if you could share my LinkedIn, um, so if you could just type on my name. So definitely you would find me and we can, can get connected. And I would really appreciate, um, Ann and the Everyday PM podcast for enabling me to share some of the, uh, best practices knowledge, exposure. So it's more of like conversational kind of topics and I truly enjoy and anyone who wants to connect with me and learn more about any sort of collaboration. Yes, I'm open to it.

Speaker A: It's been, yeah, truly an honor and a pleasure hosting you here. I'm very excited to release these again as a series, as Dr. Malad mentioned. A good conversation to listen to from start to finish in these episodes. So thank you so much for joining us. If you'd like to continue the conversation with me, you can find me on LinkedIn as well. I'll drop that link into the, uh, podcast description. Make sure to follow and subscribe to the Everyday PM podcast. You can find it on every podcasting platform and let us know what you thought about today's episode. So thank you so much for joining us today and for the important work you're doing to ensure AI serves the construction industry responsibly. Dr. Mala, first and foremost, thank you for that. And to our listeners, as you explore AI in your own projects, remember, technology is a tool, but ethics is a choice, so make it intentional. Intentional. And thanks for listening to our episode. And until next time, take care.