DATA AND AI ARE CHANGING THE WAY ORGANIZATIONS THINK, DECIDE, AND ORGANIZE. IT’S TIME HUMANITIES, MANAGEMENT AND SOCIAL SCIENCES GET INVOLVED.
Metalab

NEWS

Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat. Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur. Excepteur sint occaecat cupidatat non proident, sunt in culpa qui officia deserunt mollit anim id est laborum.

EVENTS

[ARTICLE] This paper reviews bilevel optimization, which models hierarchical decision processes and is challenging to solve.

by Ivana Ljubić (ESSEC Business School), Thomas Kleinert, Martine Labbé, Martin Schmidt

Bilevel optimization is a field of mathematical programming in which some variables are constrained to be the solution of another optimization problem. As a consequence, bilevel optimization is able to model hierarchical decision processes. This is appealing for modeling real-world problems, but it also makes the resulting optimization models hard to solve in theory and practice. The scientific interest in computational bilevel optimization increased a lot over the last decade and is still growing. Independent of whether the bilevel problem itself contains integer variables or not, many state-of-the-art solution approaches for bilevel optimization make use of techniques that originate from mixed-integer programming. These techniques include branch-and-bound methods, cutting planes and, thus, branch-and-cut approaches, or problem-specific decomposition methods. In this survey article, we review bilevel-tailored approaches that exploit these mixed-integer programming techniques to solve bilevel optimization problems. To this end, we first consider bilevel problems with convex or, in particular, linear lower-level problems. The discussed solution methods in this field stem from original works from the 1980’s but, on the other hand, are still actively researched today. Second, we review modern algorithmic approaches to solve mixed-integer bilevel problems that contain integrality constraints in the lower level. Moreover, we also briefly discuss the area of mixed-integer nonlinear bilevel problems. Third, we devote some attention to more specific fields such as pricing or interdiction models that genuinely contain bilinear and thus nonconvex aspects. Finally, we sketch a list of open questions from the areas of algorithmic and computational bilevel optimization, which may lead to interesting future research that will further propel this fascinating and active field of research.

[Please read the research paper here]

[ARTICLE] An integer linear programming-based solution strategy has been developed for leaf sweeping operations in Trenque Lauquen, Argentina, optimizing sweeper assignments, leaf bag deposit points, and collection truck routes.

by Diego Delle Donne (ESSEC Business School), Valeria Di Tomaso, Guillermo Duran

A solution strategy based on integer linear programming models has been developed for leaf sweeping operations in the Argentine city of Trenque Lauquen. The aim is to achieve efficiency in the assignment of sweepers to city blocks, the identification of leaf bag deposit points and the routes to be followed by collection trucks for leaf bag pickup. Previous to this strategy, sweeper assignments were improvised and inefficient, with blocks often left unswept. Furthermore, no method was available for accurately determining the number of sweepers needed to ensure either full coverage of all city zones within the working day or a balanced work load distribution across all sweepers. Application of the solution strategy by the city has resulted in efficient definitions of sweeper requirements while optimizing sweeper assignments such that all blocks are covered. Once the strategy is fully implemented, the number of bag deposit points under the manual definitions should be reduced by roughly one-half and the total travel distance of the truck routes, modelled as an asymmetric travelling salesman problem, should be cut by 10–15% with the consequent savings in time, vehicle use and fuel consumption.

[Please read the research paper here]

[ARTICLE] This paper studies the Minimum Sum Coloring Problem (MSCP) and introduces the first branch-and-price algorithm to solve it to proven optimality.

by Diego Delle Donne (ESSEC Business School), Fabio Furini, Enrico Malaguti, Roberto Wolfler Calvo

A proper coloring of a given graph is an assignment of a positive integer number (color) to each vertex such that two adjacent vertices receive different colors. This paper studies the Minimum Sum Coloring Problem (MSCP), which asks for finding a proper coloring while minimizing the sum of the colors assigned to the vertices. We propose the first branch-and-price algorithm to solve the MSCP to proven optimality. The newly developed exact approach is based on an Integer Programming (IP) formulation with an exponential number of variables which is tackled by column generation. We present extensive computational experiments, on synthetic and benchmark DIMACS graphs from the literature, to compare the performance of our newly developed branch-and-price algorithm against three compact IP formulations. On synthetic graphs, our algorithm outperforms the compact formulations in terms of: (i) number of solved instances, (ii) running times and (iii) exit gaps obtained when optimality is not achieved. For the DIMACS instances, our algorithm is competitive with the best compact formulation and provides very strong dual bounds.

[Please read the research paper here]

[ARTICLE] This paper addresses the Double Traveling Salesman Problem with Multiple Stacks (DTSPMS), focusing on its computational aspects, demonstrating polynomial complexity under certain conditions for key subproblems.

by Laurent Alfandari (ESSEC Business School), Sophie Toulouse

The Double Traveling Salesman Problem with Multiple Stacks, DTSPMS, deals with the collect and delivery of n commodities in two distinct cities, where the pickup and the delivery tours are related by LIFO constraints. During the pickup tour, commodities are loaded into a container of k rows, or stacks, with capacity c. This paper focuses on computational aspects of the DTSPMS, which is NP-hard. We first review the complexity of two critical subproblems: deciding whether a given pair of pickup and delivery tours is feasible and, given a loading plan, finding an optimal pair of pickup and delivery tours, are both polynomial under some conditions on k and c. We then prove a (3k)/2 standard approximation for the Min Metric k DTSPMS, where k is a universal constant, and other approximation results for various versions of the problem. We finally present a matching-based heuristic for the 2 DTSPMS, which is a special case with k=2 rows, when the distances are symmetric. This yields a 1/2−o(1), 3/4−o(1) and 3/2+o(1) standard approximation for respectively Max 2 DTSPMS, its restriction Max 2 DTSPMS(1,2) with distances 1 and 2, and Min 2 DTSPMS(1,2), and a 1/2−o(1) differential approximation for Min 2 DTSPMS and Max 2 DTSPMS.

[Please read the research paper here]

[ARTICLE] This paper introduces the Sparse CP-VAR model to address the parameter proliferation in Change-point VAR models.

by Arnaud Dufays, Jeroen V.K. Rombouts (ESSEC Business School), Zhuo Li, Yong Song

Change-point (CP) VAR models face a dimensionality curse due to the proliferation of parameters that arises when new breaks are detected. We introduce the Sparse CP-VAR model which determines which parameters truly vary when a break is detected. By doing so, the number of new parameters to be estimated at each regime is drastically reduced and the break dynamics becomes easier to be interpreted. The Sparse CP-VAR model disentangles the dynamics of the mean parameters and the covariance matrix. The former uses CP dynamics with shrinkage prior distributions, while the latter is driven by an infinite hidden Markov framework. An extensive simulation study is carried out to compare our approach with existing ones. We provide applications to financial and macroeconomic systems.

[Please read the research paper here]

[ESSEC Knowledge] by Veronica Casarin - Professor at ESSEC Business School

Organizations often face conflicting values and challenges in defining success. The author proposes here a new framework to tackle this.

Her research introduces "calculative frames," a model that captures different perspectives on value within an organization. This helps understand how organizations navigate conflicting priorities, like social good vs. financial sustainability.

By analyzing how these perspectives are balanced through key metrics, Veronica Casarin reveals how organizations reach compromise. This research offers valuable insights for any organization facing diverse stakeholder demands and complex value measurements.

[To read the full article please follow this link.]

THE POTENTIAL OF INTERDISCIPLINARY COLLABORATION IN AI SAFETY

[Student IDEAS] by Pieter Jan Motmans - Master in Data Sciences & Business Analytics at ESSEC Business School & CentraleSupélec

Abstract

As AI systems are getting increasingly suited for real-life applications, research on how these systems can be made safer is of the utmost importance. The field of study performing this research is called AI Safety. In this article, the focus is on methods developed within AI Safety to make AI systems uncertain over their true objective. To impute this uncertainty, current methods rely on human oversight during training. However, there has not yet been a lot of focus on the differences in outcome when the system was trained by different individuals with different preferences. The main thesis of this article is that such questions require an answer, and that great progress could be made by executing the kind of experiments which are commonplace in the social sciences. 

---

Fast forward several (being intentionally vague here) years into the future. You have just stepped into your self-driving car and tell it to bring you to work. Turns out, after an overnight update, it found a faster way to arrive at work. Without asking for your permission, the system starts speeding in pedestrian areas and takes a turn to drive in the opposite direction of the traffic flow. Maybe it will get you there faster, but in doing so the car posed a threat to pedestrians and other cars. 

I can already hear you think, telling the car to get to work as fast as possible is not a good idea. Rather, we should tell it to get there as fast as possible while abiding by certain rules. I encourage you to come up with potential rules, and then think about loopholes for them. If this goes anything like it did for me, you will soon realise that coming up with a coherent set of rules is almost impossible. Adhering to the law would be a good start, but how to tell the car that it normally can’t overtake on the right side, but can do so in case of a traffic jam? In short, trying to come up with all eventualities before deployment is, at least according to me, infeasible. 

Let’s analyse what happened here. Self-driving cars would be trained with a method called Reinforcement Learning. The concept is intuitive: the system is given positive feedback for actions that lead to good outcomes, and punished for actions that lead to bad outcomes. As such, the hope is that good actions are reinforced, and bad actions are discouraged. This way of learning closely mimics the way humans and animals learn, and is therefore often thought of as a promising approach to Artificial Intelligence. 

One of the central problems with this approach consists of trying to define what a good outcome entails. Current systems are given a reward function that specifies the desired behaviour. In the self-driving car example an initial reward could be to get to the destination as fast as possible, and after experiments show dangerous driving by the car, this reward could be changed to getting to the destination as fast as possible while abiding by certain rules. 

In comes AI Safety

The field of AI safety focuses on avoiding such unexpected and/or harmful consequences that arise when deploying AI systems in the real world. The field is broad and proposes various approaches to doing so. This article will focus on a subfield of AI Safety that aims to reasonably align AI Systems with human values: AI Alignment. Its hypothesis is that many of such potential problems share a common source. Namely that by letting the system optimise a fixed reward function, this reward function becomes the system’s only source of truth, making it impossible for humans to control the system once it has been trained. 

Current research on AI Alignment proposes to solve this issue by making the system inherently uncertain over its objective. Therefore, it will always look to humans to find the true objective. Today, most alignment research is technical, searching for different methods to transition from having a fixed reward, to making the system uncertain over its true objective. 

What several of these approaches share, is their involvement of a human guide for the AI’s training. However, it has not been well mapped how sensitive AI systems are to the background of that guide. It seems that the social sciences could help in this mapping.

Before describing how I believe this could happen, I deem it necessary to describe three of the proposed alignment methods here. First, Inverse Reinforcement Learning (IRL) aims to make AI systems learn by observing human behaviour. It is well summarised by these three principles outlined by Stuart Russel in “Human Compatible”, the first two are more generally applicable to AI Alignment as a whole, whereas the third hints at the specific method IRL proposed: 

  1. The machine's only objective is to maximize the realization of human preferences.
  2. The machine is initially uncertain about what those preferences are.
  3. The ultimate source of information about human preferences is human behavior.

Second, Reinforcement Learning from Human Feedback trains AI systems by giving feedback on their actions. This is achieved by showing a human two videos of the systems’ behaviour, and asking which one shows the better behaviour. OpenAI pushes forward this branch of research, and has used it to train Chat-GPT to not respond to questions such as “How can I harm myself?” or “How can I build a bomb?”. Finally, a third method called Debate aims to let a human judge a debate between two AI systems. A question is asked, and the two AI systems make statements in support of their answer. In the end, the human judges which system won. The reason I mention Debate is because it forms the central method in an OpenAI paper with a similar goal: attract social scientists to the field of AI Safety.

AI Safety needs social scientists

Now we come to the central thesis of this article: social scientists can advance AI Safety research. In suggesting this, I am not alone. In 2019, OpenAI published a paper called “AI Safety needs social scientists”. I believe there are several reasons to come back to the arguments made in that article, and expand on them. First, the paper focuses almost exclusively on the approach to AI Alignment called Debate, and states that similar questions can be applicable to other methods to learn about human values. With what is known today, it is important to make these similar questions explicit. Second, the goal of the paper can broadly be defined as finding good judges. That is, finding people such that when they judge the debate, the debate is successful. I do not agree with this goal. Finally, whereas in 2019 it was not possible to use an AI+AI+Human setting for the debate, the rise of large language models allows exactly this setting. Indeed, it takes away one of the main constraints encountered at the time of writing of the paper, namely that to replace the AI debaters with humans, domain experts were required to participate in experiments. That is both expensive and inefficient. Now, every social scientist at least has access to sufficiently powerful AI debaters. 

Research questions for social scientists

This section covers some areas in which I believe that social science research is needed to complement current research efforts in AI Alignment. To the best of my knowledge, research in these specific areas has not yet been performed. I encourage the reader to prove me wrong on this, and would be very happy if that were to happen. This section remains rather exploratory, giving pointers rather than complete solutions. 

Impact of the individual on the outcome

When using human feedback or behaviour to train AI systems, a first question is to what extent the specific individual has an impact on the outcome of the training. Research on psychological attributes is ample, the Big Five personality traits are an example of an attempt to objectivize personality. A classical psychological experiment could then try to identify relationships between such personality traits that the individual exhibits, and the outcomes of AI systems trained with human guidance. 

Similar experiments can also be done on the basis of cultural differences. Hofstede’s taxonomy is still used to study the differences between different cultures. If experiments can be done to search for cause and effect in cultural differences, and AI outcomes, a good start could be made to look at the cross-cultural differences in human values. 

Unlike the proposal of OpenAI to identify superjudges, people whose judgement in debates leads to better outcomes, the research proposed here simply aims to better map the effect the specific individual has on the outcome of an AI system. As such, the aim is not to put certain people in a privileged position. In fact, I would argue that doing so would be detrimental. Having a limited number of people to bring across human values can clearly not bring them across in all their richness and diversity.

To make it more tangible, we can outline how such an experiment would go for the self-driving car mentioned above. In this situation, the system could be trained on the driving behaviour of an individual. It is then not unthinkable that the system learns slightly different preferences from different individuals, as the different driving styles might point to different trade-offs of speed and safety while driving. Would the self-driving car then also drive ‘riskier’ when trained on the driving style of someone who values getting to their destination quickly?
Other examples more specific to the social sciences can be found in “Thinking, Fast and Slow”. The book describes plenty of experiments to show the different biases that lead people to make suboptimal decisions. In general, one could devise an experiment in which an individual first participates in such an experiment, before training a system on the basis of human feedback to perform the same experiment. Potentially interesting links could then be found between the individual’s performance and the AI outcome. People better versed in psychological experiments could undoubtedly come up with much more interesting experiments, and I do believe that many of such experiments would lead to interesting insights.

Multi-agent analysis

Research on multi-agent systems, in which several AI systems interact, compete or collaborate, is also well established. Game theory is a framework well suited to this study, in which the outcome arising from the interaction between rational agents is studied. Sociology is another field that aims to understand the behaviours that arise from human interaction. 

Such research can be aimed at understanding the group dynamics when multiple agents trained on human feedback or behaviour interact. This could be compared with systems not specifically trained to align with human values to assess how important technical AI alignment is. Additionally, it could be used to test how promising a certain method for AI alignment actually is. If systems trained on human feedback show desirable group dynamics, and those trained with debate don’t, it is a hint to the promise of human feedback. 

A hint in this direction is found in “Generative Agents: Interactive Simulacra of Human Behavior”. In this paper, the authors present AI Agents designed to mimic human behaviour in a Sims-like environment. These agents are shown to work, interact with each other and form ‘opinions’. These agents are largely based on GPT-technology, and therefore fully act in natural language. Additionally, it is possible for a user to give an instruction to an agent such as: “organise a party next week”. Both these factors make it easy for researchers in different areas to study these agents. The authors recognize this potential, and mention the testing of social science theories as one of the potential applications.

Three types of social behaviour were already noticed in the paper. First, there was diffusion of information, one agent was planning to organise a party, and by the end of the simulation, several agents knew about it. Second, at the end of the simulation they remembered their interactions with others and could form opinions about each other. Finally, they showed cooperation in the preparation of the aforementioned party. 

These agents are largely based on GPT-technology, and were trained with human feedback. However, more and more large language models are developed now so alternative foundations for these agents could be developed with a different focus on alignment. As such, it is potentially a good use case for the type of research we propose here, and indeed we can study the emergent social behaviour of agents based on models trained with different alignment methods. 

Challenges

Clearly, such research does not solve everything, and some challenges still arise. First, deep learning systems are so-called black boxes, we can witness how the system reacts to a certain event, but it is difficult to know why, and through what process, the system decides on its reaction. By trying to align systems with human values, by effectively teaching the system human values, we run the risk of creating an additional ‘moral black box’. In this case, the system seemingly behaves according to human values, but we are still unable to know exactly what the system learned. Even worse, the system could become manipulative. Making us believe it is well aligned and safe, but then changing its behaviour when deployed in reality. If such a thing happens, it becomes questionable whether it is even desirable to try to impute human values into AI systems. It is for this reason that AI Alignment needs to be complemented by AI Explainability. 

Second, what matters in the end for now is still human decision making, and the incentives that lead to it. Research on AI Alignment can not take away the responsibility from the people that design and use AI systems. For example, the agents we described above that imitate human behaviour could lead to a lot of misuse. This standpoint that the responsibility still lies with the human should be interpreted as beneficial and empowering, rather than something that should be avoided. 

Conclusion

With this article, I hope to have raised interest in the interesting field of AI Alignment. I believe there was a need to write it because the potential for contributions to AI Alignment by social scientists has not yet been fully acknowledged, whereas other fields of AI research already benefit from such contributions. 


By acknowledging the limitations of this research field, it becomes clear that AI Alignment should not be the only focus for AI researchers. However, it is a valuable complement to research in AI explainability and AI regulation, both of which are a focus of colleagues in Institute for AI & Business Ideas.

[ESSEC Knowledge] by Olga Klopp - Professor at ESSEC Business School

“Big data”, “high-dimensional statistics”, “predictive analytics” … while we have likely all heard at least one of these terms, their meanings are a little more elusive. How are they applied? How can advances in data analytics impact our daily lives? ESSEC professor Olga Klopp and her colleague Geneviève Robin from École des Ponts ParisTech have developed a new method of data imputation to address the challenges of modern data analysis. 

For data to be analyzed, it must first be transformed from its raw form into a more usable format. We do this in myriad ways every day: we look at a person’s face and can identify the combination of features as our colleague or friend, we hear a sound and identify it as a car horn, we smell something in the air and realize that there is a bakery nearby. We also transform data in our professional activities: a doctor will take a list of symptoms and “transform” it into a diagnosis and prognosis. For the latter to occur, we need to be able to predict: in other words, to predict something we have not observed, based on the data we have observed. Think, for example, of a patient who has been in a serious car accident. This person is at risk of experiencing hemorrhaging, which would put their life at risk. The presence of hemorrhaging thus needs to be detected early to treat the patient and possibly save their life. Doctors have thus identified a number of factors that are related to hemorrhaging: in other words, they predict a non-observed trait (the hemorrhaging) from a series of observed traits (for example, the person’s level of hemoglobin).

However, reality is a bit more complicated, as the accuracy of this prediction is tied to three hypotheses. Namely, that the observed factors are related to the variable to be predicted, that we have prior data at our disposal, and that we have a model that we can use to synthesize and interpret the data. It is not always possible to verify these hypotheses. Often we are interested in a phenomenon for which we don't yet know which indicator, trait or factor have been proven to be related to it.  For example, in the genome study of cancer, where we want to identify genes potentially linked to its development, we generally have measurements made for thousands or even millions of genes, but we don't know which ones are relevant to the type of cancer under study. This results in a problem referred to by statisticians as “dimensionality reduction”, meaning the process of reducing the vast number of variables and information available to a small, manageable number that are closely linked to the phenomenon we are interested in. 

[To read the full article please follow this link.]

[ESSEC Knowledge] by Harris Kyriakou - Professor at ESSEC Business School

Is novelty trading off with practicality? In new research forthcoming in MIS Quarterly (1), Harris Kyriakou (ESSEC Business School) and his colleagues Jeffrey Nickerson (Stevens Institute of Technology) and Ann Majchrzak (University of Southern California) explored how novelty, and the structure of preexisting designs affect product development processes in online innovation communities.  

What are online innovation communities? 

Online innovation communities, rather than focusing on participants’ profiles, focus on the development of products. They’re distinguishable from other online knowledge production communities like Wikipedia and open source software forums in three key ways:

  1. They don’t have one single production goal, with the content being evaluated based on its novelty, rather than the knowledge it includes
  2. They are designed around artifacts made by individuals, not team-based projects
  3. Novelty plays a crucial role in retaining and keeping participants engaged with the community 

In online innovation communities, the term “design landscape” is used to refer to the place where the search for a design occurs (2). An array of artifacts forms this design landscape, and members of the community search the landscape for new designs and add their own. The idea is to add and identify artifacts that are novel in relation to other artifacts: novel artifacts capture people’s attention. However, past work has also suggested that novelty creates uncertainty, leading associated costs to increase (3, 4). 

[To read the full article please follow this link.]

[ESSEC Knowledge] by Jeroen Rombouts - Professor at ESSEC Business School

Article written with Gérard Guinamand, CEO DATatSCALE

What has changed in data recently?

In recent years, companies have become increasingly aware of the potential of technological advances such as connected objects, and analytics platforms, which now function in real-time for decision-making. In addition, companies are open to ecosystems of partners including suppliers, customers, and even competitors. Jeroen Rombouts, professor of statistics and econometrics at ESSEC Business School and Chair of the Accenture Strategic Business Analytics Chair, and Gérard Guinamand, CEO of DATAtScale and former Chief Data Officer at ENGIE, share their insights.

Given our experience as Chief Data Officer and the handling of many use cases over the last five years, we identify what makes data initiatives successful. 

A common element of these advances is the role of data: generated and used at lightning speed, this data can now be sold or shared between companies. Not just that: data analysis is a huge value add for these companies. Given the significant investments involved, companies are now increasingly interested in highlighting the value generated by these data and analytics efforts. 

CEOs and company boards are now convinced that this data adds significant value to  their organization. In addition to its use for managing KPIs, senior management understands that data can be a source of revenue thanks to new products and services. Data can also be an accelerator of operational excellence by offering new performance levers: predictive maintenance of industrial assets and process automation. In addition, CEOs also see data as an opportunity to reduce commercial, industrial, and regulatory risk.

[To read the full article please follow this link.]

Founded in 2020 by ESSEC Business School, The Institute for AI and Business for Artificial Intelligence, Data and Society helps organizations navigate and better understand the social, economic, cultural, and ethical impacts of AI and data

metalab@essec.edu

Learn more about the Institute for AI and Business

copyright © 2026 Institute for AI and Business

↓
Résumé de la politique de confidentialité

Ce site utilise des cookies afin que nous puissions vous fournir la meilleure expérience utilisateur possible. Les informations sur les cookies sont stockées dans votre navigateur et remplissent des fonctions telles que vous reconnaître lorsque vous revenez sur notre site Web et aider notre équipe à comprendre les sections du site que vous trouvez les plus intéressantes et utiles.