When Working Traffic Lights Stop Working: The Paradox of Human Intervention

 By Sebaggala Richard (PhD)

Few days ago, I found myself stuck at a traffic junction in Kampala. It was not because the traffic lights had failed. It was not because there had been an accident. It was not because the road had been closed. The traffic lights were working perfectly well, changing from red to green and back again. Yet, strangely, they no longer mattered. Traffic officers had taken over the junction, stopping one side of the road for what felt like an unreasonable length of time while allowing another side to move almost continuously.

 

For those of us waiting, the frustration was not just about delay. Kampala drivers are used to delay. What made the situation more painful was the feeling that the system had become unfair. The lights kept changing, but our turn never came. One side of the road seemed to have been given priority, as if the people in the other cars had more important business than the rest of us. In that moment, the question that came to my mind was not simply why traffic was moving slowly. It was why a functioning rule-based system had been suspended and replaced with human discretion.

 

That small moment at a Kampala junction says something much bigger about how societies are organised. Traffic lights represent a rule-based system. They coordinate strangers without asking who they are, where they are going, or how important they think their journey is. When the light is red, you stop. When it is green, you move. The system may not be perfect, but it is predictable. More importantly, it is impersonal. It does not favour the powerful, the impatient, or the well-connected.

 

Traffic officers represent a different form of coordination. They rely on judgement. In some cases, this can be useful. If one road is heavily congested and another has very few vehicles, an officer may decide to give more time to the busier side in order to clear the queue and prevent wider gridlock. In principle, this is not unreasonable. A good traffic system should not always treat unequal conditions equally. Sometimes efficiency requires giving more attention to the side where congestion is worse.

 

But this is where the problem begins. Discretion only improves outcomes when it is guided by evidence, clear rules, and accountability. Without data on queue lengths, traffic volume, waiting time, and spillover effects, human judgement can easily become arbitrary. What may look like traffic management to the officer can feel like punishment to the people waiting. The officer may believe he is solving one problem, while creating another problem of frustration, mistrust, and perceived unfairness.

 

Economics helps us understand why people become angry in such situations. Human beings do not only care about outcomes; they care about how outcomes are produced. A driver may tolerate waiting if everyone else is also waiting under a clear and predictable system. But when one lane moves continuously while another is held for thirty minutes, people start comparing their treatment with that of others. The loss is no longer just time. It becomes a loss of fairness. This is why workers compare salaries, taxpayers compare tax burdens, and citizens compare access to public services. People want efficiency, but they also want procedural justice.

 

The Kampala traffic junction therefore becomes a mirror of a wider institutional challenge. In Uganda, we often have formal systems, but we frequently suspend them in favour of personal intervention. We have rules, but we also have exceptions. We have procedures, but we also have someone who can override them. We have institutions, but we often rely on individuals to make them work. The result is a society where people learn not only to follow rules, but also to look for who has the power to bend them.

 

This is not just a traffic problem. It is a development problem. Investors care about predictable regulations. Businesses care about predictable taxes. Students care about predictable academic rules. Citizens care about predictable public services. Drivers care about predictable traffic systems. In all these cases, predictability reduces the cost of decision-making. When systems are predictable, people can plan. When systems depend too much on discretion, people spend their energy trying to interpret, negotiate, or survive the mood of the person in charge.

 

The irony is that discretion is often introduced in the name of solving immediate problems. A traffic officer may override the lights because the junction appears congested. An administrator may bypass a procedure because the process appears slow. A public official may create an exception because the formal rule appears inconvenient. In the short run, this may look practical. But when exceptions become normal, rules lose authority. People stop trusting the system and start looking for personal routes around it.

 

This is one of the quiet differences between strong institutions and weak ones. Strong institutions do not depend on exceptional individuals to function every day. They are designed to work even when nobody is watching closely. A traffic light is not wise, but it is consistent. It does not get tired. It does not favour one road because someone important is coming. It does not punish one lane because the officer misjudged the queue. Its strength lies in its impersonality.

 

Of course, this does not mean traffic officers are useless. Emergencies happen. Accidents happen. Presidential convoys happen. Roads flood. Junctions become overwhelmed. In such moments, human judgement is necessary. The problem is not discretion itself. The problem is discretion without discipline. A well-managed city should know when officers may override traffic lights, for how long, under what conditions, and with what accountability. Otherwise, intervention becomes improvisation, and improvisation becomes disorder.

The deeper question is whether we are building systems that reduce the need for constant human rescue. Development is not only about constructing roads or installing traffic lights; it is about building institutions that are trusted enough to operate. If every functioning system still requires someone to stand over it, interrupt it, and reinterpret it, then the real problem is not the technology—it is institutional confidence.

However, change is possible. Just a few days ago, I went to renew my driving permit, and the efficiency with which the process was completed left me wondering, for a moment, if I was still in Uganda. It served as a powerful reminder that even in a landscape often defined by disorder, pockets of effectiveness exist. That experience demonstrated that efficient public service delivery is not just a dream—it is a tangible possibility, and it brings with it the kind of predictability and ease that every citizen deserves. I often find myself wishing that all our institutions functioned with the same clarity as our driving permit services; if they did, Uganda would truly be a different place.

Ultimately, we must recognize that the difference between order and disorder is rarely the absence of systems. Too often, it is simply the refusal to let those systems do their work.

 

When AI Meets Academic Publishing: Africa's Chance to Rewrite the Rules

By Richard Sebaggala (PhD)

 

For years, academic publishing has been one of the strangest industries in the modern economy. Universities pay researchers to conduct studies. Researchers write articles and submit them to journals. Other researchers review those articles, usually without compensation. Publishers then package the finished product and sell it back to the very universities that funded its creation.

This arrangement has survived for decades because it solved an important problem: trust. In a world flooded with ideas, someone had to decide which research deserved to be taken seriously. Journal publishers became the gatekeepers of that process. Over time, a small number of large publishing houses accumulated enormous influence over how knowledge is evaluated, disseminated and rewarded.

Artificial intelligence is beginning to unsettle that equilibrium. Much of the current discussion focuses on whether AI will help researchers write papers faster. That is the least interesting question. The more important question is what happens when the technology starts weakening the economic foundations on which the publishing industry was built.

The traditional publishing model relied on scarcity. Producing a publishable manuscript required substantial time, technical skill and access to expertise. Reviewing and editing required even more. AI is steadily reducing those costs. Tasks that once demanded weeks of effort can now be completed in hours. Literature reviews can be organised more quickly. Statistical code can be generated on demand. Language barriers can be reduced. Drafts can be improved, translated and reformatted with unprecedented speed.

Yet while the cost of producing content is falling, the cost of verifying content may be rising. This is where the economics becomes interesting. The value of a journal has never been the paper it prints. The value lies in its ability to convince readers that the paper is worth reading. In economic terms, journals function as trust-producing institutions. They help solve an information problem. Readers cannot personally verify every claim, dataset and methodology. Instead, they rely on journals to perform part of that verification on their behalf.

AI complicates that role. As it becomes easier to generate polished academic content, distinguishing between rigorous and weak scholarship becomes more difficult. The challenge facing journals is therefore not a shortage of manuscripts but a shortage of credible signals.

This is why predictions about the collapse of academic publishing should be treated cautiously. History suggests that powerful institutions rarely disappear simply because technology changes. More often, they adapt. The large publishers may find that selling access to journals becomes less profitable than selling access to AI-powered research infrastructure, manuscript screening systems, scholarly databases and analytics platforms. The source of market power may shift, but market power itself may survive.

At first glance, this might sound like good news for universities that have long struggled with subscription costs and publication fees. Many African institutions have experienced the publishing system as consumers rather than architects. If AI reduces the costs of producing and sharing knowledge, that must be a welcome development.

The prospect of lower publishing costs and greater competition sounds attractive. However, economics suggests that the relationship between competition and welfare is not always straightforward. Information markets depend heavily on trust. Lower barriers to entry can encourage innovation, but they can also make quality harder to assess. Academic publishing is not like selling maize or mobile phones. A journal does not simply distribute content; it certifies credibility. Its value rests on the confidence that readers place in its editorial standards and quality assurance processes.

 This distinction matters because AI is reducing the cost of both good and bad scholarship. Professional-looking journal websites can now be built cheaply. Editorial communication can be automated. Manuscripts can be polished rapidly. Publication processes can be accelerated. The result may be an explosion in journal publishing and research output without a corresponding increase in knowledge.

That possibility should concern universities more than the future profitability of the major publishing houses. The real opportunity for Africa does not lie in producing more journals or more papers. It lies in building stronger institutions for evaluating knowledge. For decades, many African universities have operated within a publishing ecosystem designed elsewhere. AI creates a rare opportunity to rethink that dependence.

 The most promising future is unlikely to be found at either extreme. It is neither continued dependence on a handful of global publishers nor uncontrolled expansion of journals competing for attention. Instead, it lies in creating credible, affordable, and regionally owned systems of scholarly verification. Universities could collaborate to establish strong disciplinary journals, shared editorial boards, regional peer-review networks, and common standards for quality assurance. AI can support these systems, but it cannot replace them.

 The deeper issue is incentives. Universities get the research culture they reward. If promotion systems continue to emphasise publication counts above intellectual contribution, AI will simply accelerate existing weaknesses. More papers will be produced, but not necessarily more insight. If institutions instead reward originality, methodological rigour, policy relevance, replication, and societal impact, researchers will adapt to those incentives as well.

From this perspective, the future of academic publishing is not primarily a technology story. It is an institutional story. AI is merely exposing questions that have existed for years: Who controls knowledge? Who verifies quality? Who benefits from the rewards of scholarship? And who should bear the costs?

For African universities, these questions arrive at an unusually opportune moment. The continent has long been disadvantaged by a publishing system whose rules were largely written elsewhere. As those rules come under pressure, there is an opportunity not merely to participate more fully in global scholarship but to help shape the next model.

Whether that opportunity is realised will depend less on artificial intelligence than on institutional imagination. The universities that thrive will not be those that use AI to produce the largest number of papers. They will be those that use it to build more trusted, more accessible, and more credible systems for creating and sharing knowledge.

 

When Signals Fail: Rethinking Academic Value in the Age of AI

By

By Sebaggala Richard (PhD) 

While reading a recent paper by Hiranya V. Peiris titled “Large language models are not the problem,” I found myself less interested in astronomy and more concerned with what her argument reveals about universities. She argues that the anxiety surrounding AI is misplaced. The tools are not undermining the discipline. They are exposing its internal contradictions. That insight applies far beyond astronomy. It captures the tension now visible across higher education.

Universities have responded to AI as a threat to academic integrity. Essays can be generated in seconds. Literature reviews can be assembled almost instantly. Structured arguments no longer require sustained effort. At first glance, this looks like a collapse of standards. But that conclusion rests on a weak assumption, that the outputs now being automated were reliable indicators of intellectual value. Once that assumption is examined, the problem looks less like disruption and more like a pricing error in the academic system.

For a long time, universities operated in an environment where producing academic work was costly. Writing clearly, organizing arguments, and synthesizing literature required time and effort. Because these tasks were difficult, they became proxies for thinking. In economic terms, they functioned as signals. A structured essay signaled competence. A detailed literature review signaled engagement. Methodological compliance signaled rigor. These signals were not perfect measures of understanding, but they were observable and scarce, which made them useful.

AI removes that scarcity. What was once costly is now cheap. When the cost of producing a signal falls close to zero, the signal loses value. This is a standard economic adjustment. When something becomes abundant, its price falls. The same logic applies here. If any student can produce a polished essay, then the essay no longer provides reliable information about understanding. What is changing is not the existence of academic work, but the meaning of the outputs we use to judge it.

This creates an uncomfortable question. If the signals no longer tell us much, what have we been measuring? The answer is not flattering. Much of academic success has depended on the ability to reproduce recognizable patterns under constraint. Once the constraint disappears, the pattern remains, but its significance weakens. AI does not remove essays or literature reviews. It reveals their limits as indicators of real thinking.

Once this is clear, the next step follows logically. Value shifts. In economics, when one margin becomes commoditized, value moves to what remains scarce. In knowledge work, that scarce element is judgment. This is not casual opinion. It is the disciplined ability to make decisions under uncertainty. It includes choosing relevant questions, selecting appropriate methods, interpreting results carefully, and connecting findings to real-world contexts. These are not easily standardized tasks. They depend on experience, context, and responsibility.

The problem is that universities are not designed to evaluate this easily. They are built around the old signaling system. Assessment methods favor outputs that can be compared, graded, and scaled. This makes administrative sense. It reduces costs and ensures consistency. But it creates a familiar incentive problem. Institutions want evidence of learning, while students respond to what is rewarded. If structured outputs are rewarded, students will produce them, regardless of whether deep understanding is present.

AI strengthens this dynamic. It gives students a tool that fits perfectly within the existing incentive structure. What we observe is not primarily dishonesty. It is rational behavior. Students are responding to incentives as expected. This is why attempts to solve the issue through detection or restriction are unlikely to work. Monitoring tools do not change incentives. As long as the system rewards outputs that can be easily generated, those outputs will continue to dominate.

Changing this requires altering what is valued. That is where resistance emerges. Evaluating judgment is harder than evaluating structure. It takes more time, requires more engagement, and introduces subjectivity. It demands more from lecturers and reduces the efficiency of large systems. Institutions resist not because the change is incorrect, but because it is costly.

There is also a psychological dimension. The old system was predictable. Follow the structure and you succeed. A system based on judgment is less certain. Outcomes are harder to anticipate. From a behavioral perspective, this creates discomfort and encourages people to defend familiar practices. The attachment is not to quality, but to predictability.

This helps explain the current anxiety. It is not only about technology. It reflects unease with losing a system that was understood. AI has not removed the core of academic work. It has exposed that the core lies elsewhere. The real adjustment is conceptual, not technical.

Redefining academic value means shifting attention. The focus moves from producing answers to asking better questions, from summarizing content to interpreting it, from following methods to designing them, and from demonstrating effort to demonstrating understanding. Learning becomes more interactive and applied because judgment develops through engagement, not repetition.

At a deeper level, this is about intellectual honesty. AI acts as a mirror. It reflects the structure of academic systems with clarity. The reflection is uncomfortable. It shows a system that has often equated visible effort with intellectual depth and procedural compliance with genuine understanding.

A mirror does not create flaws. It reveals them. The risk is not that AI will weaken universities, but that universities will ignore what they are seeing. If institutions defend outdated measures, the gap between education and real value will continue to grow. If they respond honestly, this moment becomes an opportunity for correction.

The question is no longer whether AI belongs in the university. The question is whether the university is willing to redefine value in a world where the cost of producing knowledge has changed.

 

AI Is Not Killing Creativity. It Is Changing Who Gets to Be Creative

By

Sebaggala Richard (PhD)

 

It has become common to say that artificial intelligence is killing human creativity. You hear it in universities, in the media, and increasingly in professional spaces where people write, design, teach, or make things for a living. The fear is easy to understand. If a machine can generate an image, draft an essay, suggest a storyline, or mimic a style in seconds, then it can seem as though something deeply human is being pushed aside. Yet I think this debate often begins from the wrong starting point. The real issue is not whether AI can be creative in some abstract sense, nor simply whether it can outperform humans on selected tasks. The more interesting question is what happens when a tool makes creative production easier for ordinary people who may not have had the time, training, confidence, or support to do as much before.

Part of the confusion comes from the way creativity is discussed in public. It is often treated as though it were an all-or-nothing trait: either one is creative or one is not, either AI has creativity or it does not. But that is not how creativity works in practice. Creativity is unevenly distributed, like many other human capabilities. A small number of people operate at a very high level and produce work that surprises, unsettles, and sometimes shifts the direction of a field. Most people sit somewhere below that frontier. They may still have ideas, but they often struggle to develop them, express them clearly, or refine them into something useful. From that perspective, the arrival of AI does not necessarily mean that human creativity is being erased. It may instead mean that the baseline is shifting.

That possibility is becoming harder to ignore. Some recent studies suggest that large language models can outperform the average human on certain standardized creativity tasks. They can also produce short written pieces that judges rate quite highly in terms of novelty or quality. That matters, but it is only half the story. The same body of work also shows that the most creative human participants still tend to do better. So the evidence does not really support the dramatic claim that machines have overtaken human creativity. What it suggests instead is something more uneven: AI seems capable of lifting performance in the middle of the distribution without necessarily displacing excellence at the top.

This is a far more interesting result than either utopian or apocalyptic accounts allow. It suggests that AI may matter less as a replacement for great creators and more as a support tool for everyone else. For many people, the hardest part of creative work is not brilliance itself. It is getting started. It is producing the first draft, finding an angle, trying out alternatives, or pushing through the awkward stage where an idea still feels half-formed. These early stages are costly. They require time, effort, and a willingness to tolerate failure. AI lowers some of those costs. It helps people begin, and sometimes that is what matters most.

Economists should immediately recognize the logic here. When the cost of participating in an activity falls, more people are able to take part. That does not mean quality at the frontier disappears. It means the average level of participation and output can rise, even while the best performers remain distinct. We have seen versions of this before. Calculators did not end mathematics. Word processors did not end writing. Statistical software did not end serious empirical work. In each case, a tool removed some of the friction that made participation harder. The best people still stood out, but many more people could now do competent work than before. AI may be doing something similar for creativity.

That does not mean the concerns are misplaced. There is good reason to worry that AI-generated outputs may start to sound alike, look alike, or flatten important differences in voice and style. There is also a real danger that people may rely on these systems too heavily and lose the patience required for deeper thinking. In some settings, especially where judgment is weak, AI may produce work that appears polished while remaining shallow. So this is not a simple story of progress. It is a story of gains mixed with risks.

Still, the strongest criticism of AI and creativity often assumes too much. It assumes that if machines can help with creative work, then human originality must necessarily decline. That conclusion does not follow automatically. A tool can improve average performance without replacing the highest forms of human achievement. In fact, one of the more striking patterns in recent evidence is that AI often seems to help weaker performers more than stronger ones. That point matters because it suggests the technology may be compressing part of the gap between those who already know how to express ideas well and those who do not.

For education, this matters a great deal. In many contexts, the biggest barrier is not lack of intelligence but lack of support. Students may have ideas but struggle to structure them. Young researchers may know what they want to study but fail to turn a rough interest into a coherent question. Professionals may have genuine insight but lack the time to write clearly and quickly. In such situations, AI may not create originality out of nowhere, but it can help convert weak beginnings into usable drafts, and that is not a trivial contribution.

The same may be true more broadly in lower-resource environments. In much of Africa, for example, the issue is often not that people lack imagination. It is that many work without the layers of institutional support that help refine raw ability into polished output. Good editors are scarce. Mentorship is uneven. Feedback is delayed. Access to high-quality learning materials is limited. Under those conditions, a tool that reduces the cost of drafting, revising, or exploring ideas may have wider effects than critics in richer contexts fully appreciate. Even here, however, caution is necessary. Access to a tool is not the same as capability. AI does not automatically democratize creativity simply because it is available. People still need judgment. They still need subject knowledge. They still need some ability to distinguish a promising idea from a plausible-sounding bad one. Without that, the technology can mislead just as easily as it can help.

So the choice is not between saying that AI is destroying creativity and saying that it is liberating creativity. Both claims are too blunt. The better view is that AI is changing the structure of creative work. It is making some parts easier, broadening some forms of participation, and making some kinds of output more accessible. At the same time, it may also be encouraging sameness, overconfidence, and a false impression of depth.

The real question, then, is not whether AI can replace the best human creators. At least for now, that is the wrong benchmark. The more useful question is whether AI is raising the floor for creative work, and what follows when more people are able to produce, express, and refine ideas than before. If that is what is happening, then the social significance of AI may lie less in producing masterpieces and more in widening participation. It may not remove the frontier of human creativity, but it may change who gets close enough to it to matter. That is not the death of creativity. It is a shift in its distribution.

 

 You Do Not Need Every AI Tool: A Lesson from Econometrics

By Sebaggala Richard (PhD)

 

 

Every few weeks a new artificial intelligence tool is introduced with the promise of transforming research, teaching, writing, coding, and analysis in academia. The pace of innovation is impressive, but it has also created a certain level of anxiety within universities. Students feel compelled to experiment with every new platform they encounter. Lecturers worry about keeping pace with rapidly changing technologies. Researchers sometimes feel that failure to adopt the latest tool may leave them behind.

 

The real problem, however, is not the abundance of AI tools. It is that in trying to use all of them, researchers risk fragmenting their attention and weakening the depth of their thinking.

 

In many cases, the response to this environment has been predictable. Instead of building deep competence with a small number of tools, people begin to accumulate platforms. They open accounts on multiple systems, experiment briefly with each one, and then move to the next new tool when it appears. The result is often the opposite of what they intended. Productivity declines rather than improves.

 

Whenever I observe this pattern today, I remember a lesson from my econometrics training many years ago. At the time, we were being introduced to statistical software packages such as Stata, EViews, and SPSS. These programs were widely used in universities and research institutions around the world, and for students beginning to learn applied econometrics the choice of software seemed overwhelming. Many of us were unsure which package we should invest time in learning.

 

Our lecturer offered a simple but memorable analogy. He told us that one does not need to drive every car in order to become a good driver. What matters is learning one vehicle thoroughly and understanding how it works. He then advised us that if we learned Stata properly, we would not miss much from the other packages, and that the skills we acquired would make it easier to understand any other software we might encounter later.

 

At the time the comment seemed like practical advice about software. With experience, however, it becomes clear that the point was much deeper. The lesson was about mastery and focus. In economics we often think about the efficient allocation of scarce resources. Attention is one of those scarce resources. When attention is spread across too many tools, the quality of learning and productivity declines.

 

The current environment of artificial intelligence tools presents a similar challenge. A growing number of platforms now offer support for academic tasks such as summarizing literature, drafting text, generating code, analysing documents, and organizing research materials. Systems such as ChatGPT, Gemini, Claude, Perplexity, Elicit, Avidnote, ResearchRabbit, Scite, and NotebookLM have become increasingly visible in academic discussions. Each claims to provide significant advantages for research and knowledge work.

 

Students therefore frequently ask which of these tools they should learn. The question resembles the one we asked about econometrics software years ago. The answer is also similar. Researchers do not need to learn every available platform. What matters is developing a deep understanding of a small number of tools and learning how to use them effectively in an intellectual workflow.

 

When researchers attempt to use every tool available, several difficulties tend to emerge. The first is fragmentation of workflow. Instead of concentrating on the research problem itself, the researcher spends time switching between multiple systems. The second is superficial knowledge. Individuals may become familiar with the basic interface of several platforms without developing the skill required to use any of them effectively. The third is cognitive overload. Mental effort is directed toward managing tools rather than analysing data, developing arguments, or interpreting results.

 

There is, however, a deeper and less visible cost. When researchers constantly switch between AI systems, they do not only fragment their workflow; they also fragment their thinking. Each system structures responses differently, suggests particular framings, and nudges users toward specific ways of expressing ideas. Over time, this can weaken intellectual coherence. Instead of developing a consistent analytical voice, the researcher begins to adapt to the logic of whichever tool is being used at the moment.

 

Part of this confusion is reinforced by the strong competition currently taking place among major artificial intelligence developers. Large technology firms are investing heavily in AI systems and competing to build the most capable digital assistants. This has produced intense comparisons between leading platforms such as ChatGPT, Claude, and Gemini. Each system has particular strengths. Some are particularly effective in analysing long documents, others integrate well with search engines or cloud services, and others perform strongly in coding and structured analysis.

 

For most academic researchers, however, the differences between these systems are less important than the discussion surrounding them might suggest. Modern AI models already possess capabilities that would have been considered remarkable only a few years ago. They can summarize academic papers, assist in structuring literature reviews, explain theoretical frameworks, generate programming scripts, and help refine academic writing. The critical issue is therefore not access to artificial intelligence tools but the ability to use them thoughtfully.

 

From an economic perspective, this behaviour reflects classic problems of bounded rationality and switching costs. Each new tool requires time to learn, cognitive effort to integrate, and attention to evaluate. When these costs are ignored, researchers over-invest in exploration and under-invest in mastery. The result is diminishing returns to additional tools and, in many cases, a decline in overall productivity.

 

In practice, a focused combination of tools can already provide substantial support for academic work. Systems such as ChatGPT are particularly useful as intellectual companions during the research process. They can assist in refining research questions, clarifying conceptual frameworks, designing surveys, interpreting statistical output, and improving the structure of academic writing. When used carefully, such systems function less as automated generators of text and more as conversational partners that help researchers examine their reasoning.

 

Other platforms offer strengths in areas such as document analysis and information synthesis. Systems like Gemini are often helpful when researchers are working with large reports or multiple documents that need to be summarized and compared. Tools such as Claude have become known for their ability to handle very long texts and produce structured explanations. When used selectively, these capabilities can significantly reduce the time required to extract insights from extensive material.

 

The broader principle underlying this discussion is familiar in economics. Productivity does not necessarily increase with the number of technologies employed. It increases when individuals develop comparative advantage in the use of particular tools. A researcher who understands three systems deeply will usually work more efficiently than someone who attempts to use ten different platforms at once. Mastery compounds over time. Once the logic of AI interaction is understood, adapting to new tools becomes relatively straightforward.

 

This observation also has implications for universities. Institutions sometimes respond to technological change by attempting to introduce students to a large number of platforms. A more effective approach would focus on teaching core competencies. Students should learn the principles of AI literacy, critical engagement with algorithmic outputs, responsible and ethical use of AI, and disciplined integration of a few tools into their research workflow. The objective is not simply to familiarize students with technology but to help them think more effectively in an environment where intelligent systems are widely available.

 

Looking back, the lesson from my econometrics lecturer was not primarily about statistical software. It was about maintaining focus in a world that constantly presents new options. That insight remains highly relevant today. Artificial intelligence tools will continue to appear at a rapid pace, and debates about which system is superior will likely persist.

 

In the age of artificial intelligence, the constraint is no longer access to tools. It is the ability to think clearly while using them. The danger is not missing out on AI tools; it is becoming cognitively shallow while using them. Researchers who benefit most from these technologies will not be those who pursue every new platform, but those who develop disciplined, focused, and reflective ways of working with a few powerful tools.

 

In the same way that learning to drive one vehicle well provides the foundation for driving many others, mastering a few tools can provide the foundation for productive research in the age of artificial intelligence.

 

 

Sequencing or Stagnation? Rethinking Africa’s Artificial Intelligence Strategy

By Sebaggala Richard (PhD)

 

 

In a recent Brookings commentary titled “Why Africa Should Sequence, Not Rush Into AI,” Mark-Alexandre Doumba argues that Africa’s greatest risk is not missing the AI revolution but joining it too early. Drawing on the work of Ricardo Hausmann and Dani Rodrik, the article cautions against what it describes as premature automation. The central concern is that without adequate digital infrastructure, data governance frameworks, and productive capabilities, rapid adoption of artificial intelligence could deepen dependency rather than accelerate structural transformation. It is a thoughtful intervention in an important policy debate and one that deserves serious engagement.

At the same time, the analogy underpinning the sequencing argument merits closer examination.  The thesis implicitly treats artificial intelligence as comparable to earlier industrial technologies such as factories, heavy manufacturing, or large-scale power infrastructure. In those historical periods, countries needed to accumulate domestic skills, supply chains, and institutional capacity before industrial investment could generate sustained returns. Where this foundation was weak, industrialization often produced enclaves with limited linkages to the broader economy.

 

Artificial intelligence operates in a different space. It does not primarily reorganize physical production; it reshapes how thinking and problem-solving are organized. It influences how research is conducted, how policies are drafted, how code is written, how diagnoses are made, and how information is processed. Its deployment is largely cloud-based and does not depend on ownership of heavy physical capital. More importantly, the use of AI tools itself contributes to skill formation. Individuals often develop competence through interaction, experimentation, and repeated application. Capability therefore grows partly through adoption rather than entirely before it.

 

This reality complicates the historical logic of waiting until foundations are fully consolidated. In earlier industrial waves, late entry sometimes allowed countries to observe pioneers, import mature technologies, and expand cautiously. In the current environment, the capability frontier moves quickly and continuously. Early adopters refine processes, accumulate institutional experience, and embed AI deeply into their systems. As experience compounds, catching up becomes more demanding.

 

The pattern is already visible at the individual level. Professionals who dismissed AI tools a few years ago often find that peers who experimented early have reorganized how they conduct research, prepare lectures, analyze data, and manage projects. The difference is not limited to marginal efficiency gains. It reflects changes in workflow, iteration speed, and analytical depth. When such shifts scale across institutions and economies, divergence becomes structural.

 

The labor market concerns raised in the sequencing argument are understandable. Automation can displace certain categories of routine work, particularly in service sectors. Yet many African economies have not developed large-scale industrial employment bases comparable to those that powered earlier development trajectories elsewhere. Informality remains widespread, and productivity gaps persist. In this context, the more pressing risk may not be premature deindustrialization but the failure to cultivate high-productivity knowledge and service sectors capable of absorbing a growing youth population.

 

Artificial intelligence should therefore be viewed not only as an automation technology but also as a productivity-enhancing instrument. It can strengthen agricultural advisory systems, support diagnostic processes in health care, enhance educational personalization, improve logistics coordination, and assist public administration. In environments where documentation remains paper-based and data fragmented; AI-assisted digitization and analysis can accelerate institutional modernization. In that sense, AI can contribute to building the very foundations that sequencing advocates consider prerequisites.

 

The concern about digital dependency is historically grounded. Africa’s experience with extractive development shows how exporting raw inputs while importing high-value outputs can entrench structural imbalances. A digital parallel could emerge if data is generated locally while algorithms, platforms, and standards are designed and controlled elsewhere.

 

However, dependency does not arise solely from early adoption. It can also result from disengagement. Global AI platforms will continue to expand regardless of cautious national strategies. Data ecosystems will evolve. Technical standards will consolidate. Countries that actively cultivate domestic competence are better positioned to negotiate terms, influence governance frameworks, and adapt systems to local realities. Sovereignty in the digital age depends not only on regulation but also on participation and expertise.

 

The labor dimension is equally nuanced. The relevant comparison is not between African workers and machines in isolation, but between workers who use AI effectively and those who do not. In global service markets, AI literacy is rapidly becoming a baseline expectation. Youth who master these tools strengthen their competitiveness in remote work, digital entrepreneurship, research support, and creative industries. Delayed exposure risks widening skill gaps that become increasingly difficult to close.

 

None of this diminishes the importance of governance, infrastructure, and regulatory design. Data protection regimes, interoperability standards, and digital public infrastructure remain essential pillars of a sustainable AI ecosystem. The question is whether these frameworks must be fully consolidated before meaningful adoption begins, or whether they can evolve alongside practical engagement. Institutional learning is often iterative. Policymakers refine regulatory approaches through exposure to real-world applications and emerging risks.

 

The strategic issue, then, is not whether Africa should move early or late. It is whether it will build the capacity to shape how AI is integrated into its economies and institutions. Artificial intelligence functions as a general-purpose technology that reshapes the production of knowledge and decision-making. Countries that embed it thoughtfully in education systems, research environments, entrepreneurial ecosystems, and public administration may realize productivity gains that conventional development models underestimate.

 

The debate should not be reduced to speed versus sequencing. It should focus on whether Africa approaches AI as a passive consumer or as an active capability builder. Postponement may appear prudent, but in a rapidly evolving technological landscape it carries opportunity costs that accumulate quietly yet persistently.

 

In this context, delay is not merely caution. It is a strategic position whose consequences deserve careful reflection.

 

The Architect, Not the Builder: Preserving Scholarly Judgment in the Age of AI

By Sebaggala Richard (PhD)

Last week I spoke to a group of researchers and PhD students about artificial intelligence in scholarly writing and literature review. The mood in the room was not defensive; most participants accepted that AI tools are already reshaping academic work, and the discussion was marked by curiosity and cautious optimism. Beneath that enthusiasm, however, lay a quieter concern that went beyond plagiarism or hallucinations. What unsettled many was a more fundamental question: in embracing AI, might we gradually outsource the habits of thinking that define scholarship?

 

This concern deserves serious attention because the central risk is not misconduct but the slow erosion of intellectual ownership. For doctoral students and early-career researchers, research is not simply the production of text; it is the development of judgment. It requires working through ambiguity, weighing competing explanations, and refining arguments until they can withstand scrutiny. Large language models make many parts of this process faster by summarizing articles, suggesting theoretical connections, and interpreting statistical output with impressive fluency. The results often look polished, yet polish should not be confused with understanding.

 

During the training, I demonstrated how AI can assist with drafting search strings, organizing literature into themes, suggesting model specifications, and clarifying the presentation of regression results. The tools proved useful, but throughout the session I emphasized that acceleration does not alter the underlying logic of research. A literature review still begins with a clearly defined question and proceeds through a transparent search strategy, systematic screening, careful comparison of findings, and verification of sources. While AI can help structure these steps, it cannot determine what counts as relevant evidence or where the conceptual gap lies. Those judgments remain the responsibility of the researcher.

 

The same boundary becomes even more important in empirical work. In our example using survey data, AI was permitted to suggest possible dependent and independent variables, outline potential models, and draft statistical syntax. It could recommend robustness checks and help structure the results section. It did not, however, choose the identification strategy, justify causal claims, test assumptions, or determine the substantive meaning of the findings in context. Model choice requires theoretical grounding, causal inference demands methodological reasoning, and interpretation depends on domain knowledge. Delegating these decisions would weaken the integrity of the research.

 

Responsible use therefore begins with clarity about where assistance ends and authorship begins. Before turning to AI, researchers would do well to ask not whether its use is permitted but whether it enhances their reasoning or replaces it. There is a meaningful difference between asking AI to critique a draft and asking it to write the draft itself, just as there is a difference between using it to uncover blind spots and using it to construct an argument from scratch. Although these approaches may appear similar from the outside, they cultivate very different intellectual habits.

 

The discussion also revealed a broader cultural dimension, particularly relevant in many African academic settings where struggle is often equated with learning and difficulty is treated as evidence of seriousness. When processes become faster or more efficient, suspicion sometimes follows, as if reduced effort necessarily implies reduced rigor. AI unsettles this assumption. The ability to map literature more efficiently or clarify statistical syntax quickly does not automatically diminish depth or weaken econometric understanding. Hardship is not a prerequisite for rigor.

 

Struggle has value when it produces insight, but it adds little when it is purely mechanical. Manually formatting references does not deepen theoretical reasoning, nor does repeating routine coding steps automatically strengthen econometric judgment. Spending hours constructing search strings does not guarantee conceptual clarity. Some forms of difficulty are intellectually formative, while others persist simply because they have long been part of academic practice. The aim is not to preserve difficulty for its own sake but to preserve active and disciplined thinking.

 

In practice, thoughtful use of AI can strengthen learning. During the workshop, once some mechanical aspects of literature searching were streamlined, participants were able to devote more attention to substantive questions, such as why findings differed across contexts, where theoretical tensions remained unresolved, and how to sharpen the articulation of their research gaps. Automation, in this sense, freed cognitive space for higher-level analysis. A similar pattern emerged in empirical writing, where AI’s suggestions about alternative specifications or potential weaknesses created room to focus more carefully on identification, assumptions, and interpretation, leaving the intellectual core of the exercise intact.

 

A constructive approach is therefore to think independently first by framing the research problem, interpreting results on one’s own, and sketching the structure of the argument without assistance. AI can then be used to expand and test that thinking by identifying weaknesses, proposing alternative explanations, or improving clarity. The final step requires taking full responsibility for the work by verifying every citation, checking every claim, and ensuring that the argument reflects genuine understanding. A simple test helps clarify ownership: if AI were unavailable, could you defend your research question, theoretical framework, model specification, identification strategy, and interpretation of findings? If the answer is yes, automation has supported the work without undermining it; if not, further reflection is required before it can be considered truly your own.

 

A doctoral degree is not a document production exercise but a process of intellectual formation. AI can make writing more efficient, yet it cannot substitute for judgment. History provides perspective: calculators, statistical software, and digital databases were all met with resistance when first introduced, each innovation reducing effort in certain tasks and prompting concerns about declining standards. Research did not deteriorate; it evolved, shaped less by the technology itself than by the norms governing its use.

 

AI does not eliminate the need for careful thinking; it reduces some of the mechanical burdens that surround it. Whether scholarship becomes more superficial or more sophisticated in this environment will depend less on the capability of AI and more on the discipline of those who use it. Before generating text, it is worth pausing to ask whether the tool is being used to deepen reasoning or to bypass it. Responsible use is not about preserving hardship but about preserving judgment, and judgment remains, as it always has been, a human responsibility.