BAD DATA CANNOT BE FIXED BY A GOOD ALGORITHM: WHY DATA QUALITY WILL DETERMINE GHANA'S AI FUTURE

 INTRODUCTION

Ghana's National Artificial Intelligence Strategy (2025–2035) reflects an ambitious vision for the country's digital future. It seeks to harness artificial intelligence to drive economic growth, improve public service delivery, stimulate innovation, and position Ghana as a leading participant in Africa's emerging AI ecosystem. The strategy rightly recognises that achieving these objectives will require investments in skills, infrastructure, research, governance, and institutional capacity.

Among its most important insights is the recognition that data is a strategic asset. The strategy places considerable emphasis on data governance, digital public infrastructure, local language datasets, and the broader data ecosystems required to support AI development and deployment. This reflects an important reality: artificial intelligence cannot function without data. Every AI system, whether deployed in healthcare, agriculture, education, finance, or government, depends on data as its foundation.

Yet it is precisely at this point that a critical question emerges. Discussions about artificial intelligence often focus on algorithms, computing power, and technological innovation. While these are important considerations, they can obscure a more fundamental issue. The effectiveness of an AI system depends not only on the sophistication of the technology, but also on the quality of the information from which it learns.

This distinction matters because weaknesses in data do not remain confined to the datasets from which they originate. They are reproduced through the systems built upon them. Inaccuracies become inaccurate outputs. Historical biases become automated biases. Gaps in data become gaps in decision-making. The consequences are amplified rather than corrected.

This is why a fundamental principle has emerged across the field of artificial intelligence: bad data cannot be fixed by a good algorithm. No matter how sophisticated the model, the quality of its outputs will ultimately be constrained by the quality of its inputs.

For Ghana, this raises an important question. If data is the foundation upon which artificial intelligence is built, then attention must extend beyond the availability of data to its quality, representativeness, integrity, and governance. Whether these dimensions receive sufficient attention may ultimately determine whether the ambitions contained in the National Artificial Intelligence Strategy are realised in practice.

This article argues that while Ghana's AI strategy correctly identifies data as a strategic resource, the country's long-term success in artificial intelligence will depend less on access to algorithms and more on the quality of the data on which those algorithms are trained. In this respect, data quality is not simply a technical concern. It is a governance issue, a strategic issue, and ultimately a prerequisite for achieving Ghana's AI ambitions.

DATA, NOT ALGORITHMS

Artificial intelligence is often discussed through the language of algorithms, models, and computing power. This has contributed to a common perception that the intelligence of an AI system resides primarily in the algorithm itself. The more sophisticated the algorithm, the assumption goes, the better the outcome. While algorithms are undeniably important, this view can be misleading.

Algorithms do not create knowledge independently. They do not possess an inherent understanding of the world. Rather, they identify patterns within data and use those patterns to generate predictions, recommendations, classifications, and decisions. Their effectiveness therefore depends fundamentally on the quality of the information from which they learn. An algorithm can only work with the data it is given.

This distinction is important because it shifts attention from technology alone to the information that makes technology useful. A highly sophisticated algorithm trained on incomplete, inaccurate, or unrepresentative data may produce unreliable outcomes. Conversely, a relatively simple model trained on high-quality data may generate results that are accurate, useful, and trustworthy. The intelligence often attributed to artificial intelligence is therefore inseparable from the quality of the data that underpins it.

One of the strengths of Ghana's National Artificial Intelligence Strategy is its recognition that data is not merely a supporting resource for artificial intelligence, but a strategic asset in its own right. While public discussions about AI often focus on algorithms, applications, and technological breakthroughs, the strategy acknowledges that none of these can exist without the data ecosystems that sustain them.

This recognition is reflected throughout the strategy. Significant attention is given to data governance, digital public infrastructure, local language datasets, and the broader information environment required to support AI development and deployment. The strategy also identifies data as a critical enabler of innovation, research, public service transformation, and economic growth. In doing so, it acknowledges an important reality: the value of artificial intelligence is inseparable from the quality of the information on which it depends.

This reality is particularly important because access to AI technology is becoming less of a differentiating factor. Powerful models are increasingly available through commercial platforms, open-source initiatives, and global technology ecosystems. The challenge facing many countries is therefore not whether they can access artificial intelligence, but whether they possess the data required to use it effectively.

Particularly noteworthy in Ghana's strategy is the emphasis placed on context-specific and locally relevant data. The ambition to develop AI systems that understand Ghanaian languages, operate effectively within local contexts, and address domestic challenges recognises that artificial intelligence cannot simply be imported and expected to perform optimally. AI systems learn from the data on which they are trained. Where that data fails to reflect local realities, the usefulness and reliability of the resulting systems may be limited.

A sophisticated algorithm trained on incomplete, inaccurate, or unrepresentative data will often produce unreliable outcomes. Conversely, a relatively simple model trained on high-quality data may generate results that are more accurate, useful, and trustworthy. In practice, the effectiveness of artificial intelligence is shaped as much by the quality of its data as by the sophistication of its technology.

Indeed, Ghana's National Artificial Intelligence Strategy itself identifies insufficient local datasets that are accurate, updated, and representative as a weakness within the country's AI ecosystem, while also recognising the risks associated with AI models trained on data that is not fit for the local context.

This is why a fundamental principle has emerged across the field of artificial intelligence: bad data cannot be fixed by a good algorithm. No matter how sophisticated the model, the quality of its outputs will ultimately be constrained by the quality of its inputs.

For countries seeking to build AI-enabled economies, this distinction matters. The strategic question is no longer simply whether artificial intelligence works. In many respects, it already does. The more important question is whether the data on which it depends is sufficiently accurate, representative, and trustworthy to support meaningful decision-making. It is here that Ghana's AI ambitions encounter their most important challenge.

 

 

THE DATA CHALLENGE BEHIND GHANA'S AI AMBITIONS

The importance of data becomes even more apparent when viewed through the ambitions set out in Ghana's National Artificial Intelligence Strategy. The strategy envisions artificial intelligence supporting transformation across sectors including healthcare, agriculture, education, finance, and public administration. It also seeks to promote local language technologies, strengthen research and innovation, and expand the use of AI within government services.

As access to artificial intelligence becomes increasingly widespread, the strategic challenge for Ghana may no longer be the technology itself. Powerful AI models are becoming more accessible through global technology ecosystems, commercial platforms, and open-source initiatives. Increasingly, the more important question is whether the country possesses the quality of data required to use that technology effectively. In many respects, Ghana's AI challenge is becoming a data challenge.

Consider, for example, the aspiration to develop AI systems capable of understanding and operating within Ghana's linguistic and cultural context. Such systems require large volumes of accurate, representative, and well-structured language data. The challenge is not simply the existence of local languages, but the availability of datasets that capture their diversity, usage, and evolving forms. Without such datasets, locally relevant AI remains difficult to achieve regardless of the sophistication of the underlying technology.

The same principle applies across other sectors. AI applications in healthcare depend on accurate and consistent health records. AI systems in agriculture rely on reliable information about weather patterns, soil conditions, crop performance, and farming practices. AI-enabled public services require administrative records that are complete, interoperable, and maintained to consistent standards. In each case, the effectiveness of artificial intelligence is directly linked to the quality of the underlying data.

The challenge therefore extends beyond data availability. Large quantities of data do not automatically translate into useful datasets. Information may be fragmented across institutions, collected according to different standards, contain gaps or inconsistencies, or fail to adequately represent the populations and contexts they are intended to serve. Indeed, the Strategy's own diagnostic assessment identifies insufficient local datasets that are accurate, updated, and representative, while also highlighting the fragmentation of data across sectors and the absence of consistent practices for data collection, digitisation, and sharing.

This distinction is important because the success of Ghana's AI strategy will depend not only on technological capability, but also on the robustness of the data ecosystems supporting that capability. The challenge facing Ghana is therefore not simply one of technological readiness. It is also one of data readiness.

The implications are significant. Poor-quality data does not merely create isolated inaccuracies. Artificial intelligence can amplify those weaknesses at scale. Incomplete or inconsistent data can produce unreliable outputs. Historical patterns of exclusion or underrepresentation can be learned and reproduced by AI systems. The apparent sophistication of AI can also create a false sense of confidence, obscuring weaknesses in the underlying information on which decisions are based. The COVID-19 pandemic illustrates this principle starkly. Health systems with high-quality, representative patient data were able to deploy AI diagnostic and prediction tools rapidly. These systems could identify patterns, predict hospitalisation needs, and guide resource allocation. By contrast, health systems with fragmented patient records, incomplete demographic data, or inconsistent health metrics struggled to deploy AI effectively—not because they lacked access to sophisticated algorithms, but because their underlying data was not trustworthy enough to train and deploy AI systems safely. The bottleneck was not technological. It was data quality and governance.

 

These concerns are reflected in Ghana's National Artificial Intelligence Strategy, which recognises the risks associated with bias, underrepresentation, and weaknesses in data governance when AI systems are deployed at scale. As artificial intelligence becomes integrated into more sectors of the economy, weaknesses in data can become systemic rather than isolated. Errors are repeated more quickly. Biases affect more people. Inconsistencies are reproduced across interconnected systems. The scale that makes artificial intelligence valuable can also magnify its shortcomings.

Artificial intelligence does not eliminate weaknesses in data. It amplifies them. Where the data is strong, AI can generate significant value. Where the data is weak, the technology may simply automate and scale existing problems. The question, therefore, is not whether artificial intelligence can transform decision-making, but whether the information driving those decisions is sufficiently trustworthy to justify that transformation.

DATA GOVERNANCE IS AI GOVERNANCE

If Ghana's AI ambitions are increasingly dependent on the quality of the data on which AI systems rely, then questions about data quality inevitably become questions about governance.

Discussions about AI governance often focus on ethics, regulation, transparency, accountability, and risk management. These are important considerations. However, they can create the impression that governance begins once an AI system has been developed and deployed. In reality, governance begins much earlier.

It begins with data. Every artificial intelligence system is shaped by decisions about what data is collected, how it is collected, who collects it, how it is stored, who has access to it, and how it is used. These decisions influence not only the performance of AI systems, but also the fairness, reliability, and legitimacy of the outcomes they produce. Data governance is therefore not a separate issue from AI governance. It is its foundation.

This perspective is reflected in Ghana's National Artificial Intelligence Strategy, which identifies data access and governance as a dedicated strategic pillar, recognising that the effectiveness of AI depends on the quality, availability, and stewardship of the data on which it relies.

This distinction is important because data quality is not determined solely by technical processes. It is influenced by institutional arrangements, governance frameworks, and policy choices. Questions about data quality inevitably become questions about responsibility and control. Who determines what data is collected? What standards govern its quality? Who verifies its accuracy? Who is accountable when errors exist? And who decides whether a dataset is sufficiently representative to support decisions affecting citizens?

Data governance is not fundamentally a technology project; it is a governance and organisational discipline. Decisions about data quality, ownership, accountability, stewardship, and trust are ultimately institutional decisions that technology alone cannot resolve.

These questions are particularly relevant in the context of Ghana's AI ambitions. As artificial intelligence becomes embedded within public services, economic systems, and social institutions, the quality of data can no longer be viewed as a technical matter delegated solely to information technology professionals. It becomes a matter of public policy and governance. Decisions about data shape decisions made by AI. In this sense, the governance of data directly influences the governance of artificial intelligence itself.

The issue is not simply whether data exists, but whether it can be trusted. Trustworthy data requires more than volume. It requires standards, oversight, quality assurance, accountability mechanisms, and institutional responsibility. Without these foundations, the effectiveness of even the most advanced AI systems may be compromised.

International and continental frameworks increasingly recognise this principle. UNESCO's Data Governance Toolkit provides detailed guidance for structuring data quality and governance across the full data lifecycle—from planning and collection through processing, analysis, and use. The African Union's AI Strategy, directly relevant to Ghana as an African nation, emphasises that data quality, sovereignty, and governance are foundational to Africa's AI development. These frameworks reflect a critical consensus: data quality is not optional. It is foundational to responsible AI deployment.

If data governance is indeed foundational to AI governance, then the critical question for Ghana is not merely whether the country understands this principle. It is whether Ghana can translate this understanding into institutional practice. Recognising that data governance matters is the first step. Building the governance mechanisms that deliver trustworthy data is the next.

Viewed from this perspective, several priorities emerge. One is the need for stronger institutional arrangements to assure data quality within sectors where artificial intelligence is expected to play a significant role, particularly healthcare, agriculture, education, and finance. As AI systems become increasingly dependent on large and complex datasets, ensuring accuracy, completeness, representativeness, and consistency becomes a matter of governance rather than technical administration alone.

A second priority concerns interoperability. The value of artificial intelligence is often constrained where data remains fragmented across institutions and systems. As Ghana's digital ecosystem continues to evolve, greater attention may need to be given to the standards, governance arrangements, and accountability mechanisms that enable information to be shared and used effectively across organisational boundaries while maintaining privacy and security.

A third consideration is the question of data readiness. The effectiveness of AI systems ultimately depends on the condition of the datasets on which they are built. Understanding the strengths, weaknesses, gaps, and limitations of existing datasets may therefore become increasingly important in determining where investments in data quality and governance should be directed.

These priorities are not separate from Ghana's broader AI ambitions. They are part of the institutional foundation upon which those ambitions depend. If artificial intelligence is to generate reliable and trustworthy outcomes, the governance of data cannot remain an afterthought. It must become an integral component of AI readiness itself.

This has important implications for Ghana's broader digital transformation agenda. Over the past decade, considerable effort has been invested in digitising records, services, and transactions across both the public and private sectors. These initiatives have created valuable digital foundations. Artificial intelligence, however, introduces a different requirement. The question is no longer simply whether information has been digitised, but whether the underlying data is sufficiently accurate, complete, representative, and trustworthy to support intelligent systems.

The long-term success of Ghana's AI strategy may therefore depend less on the collection of new data and more on the institutionalisation of governance mechanisms capable of improving, maintaining, and assuring the quality of existing datasets. Artificial intelligence rewards not the quantity of data that has been digitised, but the quality of the information from which it learns.

This is why the conversation about Ghana's AI future cannot be limited to algorithms, computing infrastructure, or technological innovation. It must also address the governance arrangements that determine the quality and integrity of the data on which those systems depend. The first governance decision in any AI system is not the decision made by the algorithm. It is the decision about the data from which the algorithm learns.

FROM DIGITALISATION TO INTELLIGENCE

Over the past decade, Ghana has made significant investments in digital transformation. Public services have been digitised, digital identity systems have been introduced, payment platforms have expanded, and increasing volumes of information have been moved from paper-based processes into digital environments. These developments have rightly been regarded as important milestones in the country's digital journey.

Yet artificial intelligence introduces a different challenge. The primary objective of digitalisation is to make information accessible, searchable, and usable within digital systems. Artificial intelligence, however, operates on a different logic. It does not simply store information. It learns from it.

This distinction changes how success is measured. During the digitalisation era, the central question was whether information could be digitised and made available through digital systems. Success was often measured by the number of services digitised, the volume of records converted into digital formats, and the extent to which transactions could be conducted electronically.

The AI era introduces a different question: can the digitised information be trusted? Success is no longer determined primarily by the quantity of data available, but by its quality. Accuracy, completeness, representativeness, and governance become increasingly important because artificial intelligence relies on data not merely for storage and retrieval, but for learning, prediction, and decision-making. Digitalisation rewarded data availability. Artificial intelligence rewards data trustworthiness.

This shift has important implications for Ghana's AI ambitions. The transition from digitalisation to artificial intelligence requires a corresponding shift in priorities. The challenge is no longer simply to create digital records and systems, but to ensure that the data within those systems is capable of supporting reliable intelligence. Data that may be adequate for storage, retrieval, or transaction processing may not necessarily be adequate for training, informing, or guiding artificial intelligence.

In this sense, artificial intelligence shifts the conversation from digitalisation to intelligence. Digitalisation remains an important achievement. But in the age of AI, it becomes the starting point rather than the destination. The value of Ghana's AI ambitions will depend not only on the technologies it adopts, but on the quality of the digital foundations from which those technologies learn.

CONCLUSION

Ghana's National Artificial Intelligence Strategy reflects a clear recognition of the transformative potential of artificial intelligence. Its emphasis on innovation, digital infrastructure, research, skills development, public service transformation, and data ecosystems demonstrates a serious commitment to positioning the country for participation in the emerging AI economy. Importantly, the strategy recognises that data is not merely a supporting resource, but a strategic asset and a necessary foundation for achieving these ambitions.

Yet the long-term success of artificial intelligence will depend on more than access to technology. As AI becomes increasingly integrated into economic and social life, the more fundamental question concerns the quality of the data on which it depends. The challenge is not simply whether artificial intelligence works, but whether the information from which it learns is sufficiently accurate, representative, and trustworthy to support meaningful decisions.

Viewed from this perspective, data quality is not merely a technical concern. It is a governance concern. Decisions about how data is collected, managed, verified, shared, and governed will shape the effectiveness, fairness, and legitimacy of the AI systems built upon it. The future of artificial intelligence will therefore be influenced as much by the quality of data governance as by the sophistication of the technology itself.

For Ghana, the challenge ahead is not simply to build AI systems. It is to build the trustworthy, representative, and well-governed data foundations upon which those systems depend. Realising the ambitions contained in the National Artificial Intelligence Strategy will require recognising that the path to effective artificial intelligence begins long before an algorithm is deployed.

The next phase of Ghana’s AI journey may depend less on acquiring new technologies and more on strengthening the institutional foundations of data governance. That means improving data quality, enhancing interoperability across systems, assessing the readiness of existing datasets, and embedding accountability for data stewardship across institutions. In practical terms, Ghana must institutionalise the governance mechanisms that produce trustworthy data. Digitalisation brought Ghana into the digital age; data governance will determine whether it succeeds in the AI age. Without trustworthy data, AI simply accelerates confusion.