The Utility of Artificial Intelligence in the Pursuit of Justice through Judicial Precedent in Nigeria ()
1. Introduction
Justice Oliver Windell Holmes, a 19th Century Judge in the United States, has this say about the philosophy of law:
“The prophesies of what the Courts will do in fact, and nothing more pretentious, are what I mean by the law.”1 (Holmes, 1897)
On the other hand, Lord Edmund-Davies, a 20th Century Judge in his journal “Judicial Activism” in 1975, has this to say:
“Whatever a judge does, he will most surely have his critics. If, in an effort to do justice, he appears to make new law, there will be cries that he is overweening and that he has rendered uncertain what had long been regarded as established legal principles. On the other hand, if he sticks to the old legal rule, and equally vocal body will charge him with being reactionary, a slave to precedent, and of failing to mould the law to changing social need…” (Edmund-Davis, 1975)
The above quote by Lord Edmund-Davies, reflects the reality of judges in making laws i.e. case laws. Judges, in the process of making case laws, are bound by the common law doctrine of “judicial precedent”. Judicial precedent or legal precedent2, as a principle of law, stipulates that similar cases should be decided alike. It emphasises the necessity for maintaining certainty of the law by judges in deciding cases. Precedents abound in law reports as authority which judges must follow in deciding cases of similar facts earlier decided by superior courts. This principle of law handicaps a judge in freely exercising his judicial discretion on a point of law that has earlier been decided by a superior court of law. Because judges must follow the established authority on a decided point, judicial activism through exercise of judicial discretion is rather curtailed if not totally lost. It is a good thing for the law to be certain, but as certain as it may be, the certainty of law must be underpinned by justice (Denning, 1979: p. 293). To achieve this end, we have to examine the application of stare decisis vis-à-vis the dynamics of modernity and the role that Judges should play in developing the law and society within the framework of its certainty.
Common law is a system of law that is based on case law and precedent (The Robbins Collection, 2010: p. 5). It was developed in England in the Middle Ages (The Robbins Collection, 2010: p. 5) and was later imported into Nigeria through Ordinance No. 3 of 1863, which made applicable to Nigeria, all laws and statutes in force in England on January 1, 1900 (Akanbi, 2022: p. 25). Therefore, in terms of influence, the English common law still resonates in almost all aspects of Nigerian laws (Akanbi, 2022: p. 25) and judicial system. However, unlike the English legal system, the Nigerian legal system at present, is characterised by legal pluralism (Oba, 2013) with multiple sources of law, including common law, indigenous laws and Islamic law. The foregoing notwithstanding, English Court decisions are no longer binding in Nigeria; they are at best of persuasive effect3. Other differences between the two legal systems can be seen in the architectural structure of each judicial systems, which in our view, is reflective of the political structure of their respective territories4. Under the English Legal System, it appears the UK Supreme Court has appellate jurisdiction only and entertains appeal on points of law only (Cowie & Torrance, 2024: p. 6; Milard, 2023: p. 15). Conversely, in Nigeria, the Supreme Court has both original and appellate jurisdiction. Quite interestingly, as it stands, there is no specific legislation in the UK and Nigeria that regulates AI5 (Haie et al., 2024; White & Case, 2024).
The convergence of Artificial Intelligence (“AI”) and the judiciary is growing at rapid pace (Buckland, 2023). This is propelled by the undeniable advantages AI offers in the administration of justice. AI has been defined as the stimulation of human intelligence processes by machines, especially computer systems (Giwa & Kodjovi, 2023). The majority of systems used in the administration of justice are developed and used for administrative purposes, with the aim of generating efficiency benefits (Smuha & Hendrickx, 2023). For example, countries such as Estonia, United States of America (USA), United Kingdom (UK) and Canada have deplored different AI tools such as Salme, Harm Assessment Risk Tool (HART) and Solution Explorer into their judicial systems to effectuate a seamless administration of justice (Govindaraj, 2022).
Our discussion here, however, focuses on the introduction of AI in the judicial decision-making process. How the use of AI tools can facilitate the exercise of judicial discretion by human judges and restrict indiscretion in their use of judicial precedents, for a seamless administration of justice. It is axiomatic that law reporting is the catalogue for judicial precedent. Indeed, law reports are the tools of trade of human judges (Ayorinde, 2018). Identifying, retrieving, and analysing relevant precedents can however be an extensive and tedious task. With the emergence of AI, as a powerful tool for legal research, we interrogate how AI tools can automate this process and transform the way that human judges can discover and analyse legal precedents in their judicial decision-making process, without much ado.
2. Historical Foundations
The history of judicial precedent is traceable to the common law of England (Eboh, 2022: p. 11). In England, the history of precedent is bound up with the development of law reporting. Decisions cannot be precedents without reliable publication (Hanna, 1957: p. 8). English law reporting was divided in four periods: the time of the Year Books (1272-1537), the reports of Plowden and Coke (1537-1765), the years of the authorised reports (1765-1865) and the modern period of 1865 (Hanna, 1957: p. 8). In the Anglo-Saxon epoch, there was no evidence of any notion of judicial precedent, the codes available were only collections of judgments, illustrative of customs (Hanna, 1957: p. 8).
The development of common law of England originated after the Norman conquest of England under King Harold in 1066. The Normans with their exceptional administrative skills integrated the various customs of England which greatly influenced the English Law (Eboh, 2022: p. 11). The Norman Conquest did not bring an immediate end to the Anglo-Saxon law, but a period of colonial rule by mainly the Norman conquerors produced change (Kiralfy et al., 2024: p. 2). During this period, serious wrongs were regarded as public crimes rather than personal matters, and the perpetrators were punished by death and forfeiture (Kiralfy et al., 2024: p. 2). The government was centralised, a bureaucracy built up, and written records maintained (Kiralfy et al., 2024: p. 3).
The legal sources of the Norman period, such as Domesday Book, the Pipe Rolls, the records of the Curia Regis and various assizes, throw little light on judicial precedent (Hanna, 1957: p. 8). The Pipe Rolls were the official court records for the courts of Common Pleas and the King’s Bench (Eboh, 2022: p. 12). There were other books and/or treatise, for example the Glanvil’s treatise written in 1187 based on collection of writs, the Bracton’s Treatise, the Note Book, the Littleton’s Tenures in 1475, which cited 25 cases and was a logical study of principle (Hanna, 1957: p. 8). Also, the Doctor and Student of 1540, and the much later Year Books which were notebooks of lawyers and students, with the aim of providing materials for arguments on pleadings (Hanna, 1957: p. 8). The Year Book period cases were used as evidence of judicial tradition and not as precedents (Hanna, 1957: p. 8). Over the years, there were other publications of reports, but they were unauthorised and were much like those in the Year Books which were mere reproductions of notes taken down for the private instruction of the writer (Evans, 1946: p. 34).
Sir Edward Coke is recognised as one of the earliest judges who played significant role in the development of judicial precedent (Evans, 1946: p. 35). His writings were devoted to deducing rules from the ancient authorities, explaining and reconciling conflicting decisions and showing that the principles upon which the old cases had been decided could be applied in his day and time (Evans, 1946: p. 35). Other significant works include that of Sir John Burrow in 1765, when he published the first authorised series of decisions of particular courts in England, which approximated the modern standard (Evans, 1946: p. 34). This publication undoubtedly put the judges of England in position to develop the doctrine of stare decisis (Evans, 1946: p. 34).
In Nigeria, the history of judicial precedent is also rooted in the common law of England, because Nigeria being a former colony and protectorate of the United Kingdom, inherited the English legal system with its huge dependence on judicial precedent and law reporting (Nnamani, 2019: p. 3).
3. Philosophical Basis of Judicial Precedent
The doctrine of judicial precedent or stare decisis6 is one of the touchstones of common law legal system. The doctrine is as old as the establishment of the courts (Jain, 2017). It is derived from the legal maxim “stare decisis et non quieta movere”, meaning: “to stand by (or adhere to) decisions and not to disturb what is settled”. The core notion of this rule has been aptly described by Thomas Burns as, “when a point of law has been once solemnly and necessarily settled by the decision of a competent court, it will no longer be considered open to examination or to new ruling by the same tribunal or those which are bound to follow its adjudication.” (Burns, 1893: pp. 1-2). The doctrine of stare decisis arose from a desire for certainty and continuity in law (Anyebe, 2019: p. 21). Paton noted that, a question ought to be decided today in the same way as they were decided yesterday simply because they were decided that way yesterday (Paton, 1946). It is “at least the everyday working rule of our law” (Cardozo, 1921) which generally must be adhered to if litigants are to have faith in the even-handed administration of justice in courts (Anyebe, 2019: p. 23). Furthermore, this doctrine of law emphasises the need for there to be in place a hierarchical court system and a system of law reporting so that past cases can be studied and followed (Ikegbu et al., 2014: p. 150).
4. Guiding Principles of Judicial Precedent
4.1. Ratio Decidendi and Obiter Dictum
For purposes of the doctrine of judicial precedent, the two elements of a decision are identified as the ratio decidendi and the obiter dictum (Ikegbu et al., 2014: p. 153). Ratio decidendi means the reason behind the decision in a case7. It also means the principle of rule of law on which court decision is bounded (Garner, 2004). Obiter dictum, simpliciter, is that which is said in passing. It is the passing comment made by a judge in the course of a judgment, which may be relevant but not a direct justification for the decision (Vijayalakshmi, 2011: p. 123) in a case. Therefore, it is not every pronouncement of a judge in the course of a judgment that is a binding precedent which must be followed. It is only the ratio decidendi that is binding, the obiter dictum does not constitute a binding precedent; it is only of persuasive effect on courts. Obiter dictum may, however, receive the force of a binding precedent by the process of “subsequent adoption”.8 This is the position of the Supreme Court of Nigeria in the case of Dairo v. UBA Plc9, where per Iguh J.S.C., stated as follows:
“It is indisputable that in the judgment of Court, the legal principle formulated by that Court which is necessary in the determination of the issue in the case, that is to say, the binding part of the decision is its ratio decidendi as against the remaining parts of the judgment which merely constitute obita dicta, that is to say, what is not necessary for the decision… Where however, an obiter dictum in one case has been adopted and becomes a ratio decidendi in a later case, such obiter dictum will have to be taken to have acquired the force of a ratio decidendi and would therefore become binding…”
4.2. Per Incuriam
Per incuriam, literally translated as “through lack of care”, refers to a judgment of a court which has been decided without reference to a statutory provision or earlier judgment which would have been relevant (Thapliyal, 2016). Per incuriam does not simply mean that there was a significant inadvertence to take account of relevant authorities in deciding a case, but such failure seriously affected the reasoning in the case and therefore have affected the outcome (Emenogha, 2019: p. 123). In Nigeria, it does not lie in lower courts to question or say that a decision of superior court was reached per incuriam; they are obliged to kowtow and follow it10. Only the selfsame superior court or one that is higher than it in hierarchy, that can revisit the decision and correct it11.
4.3. Distinguishing Tools
The doctrine of judicial precedent does not permit courts to apply the ratio decidendi of a case across board and with little regard to the facts of the case before them12. Where a prior decision is presented to a court as a binding precedent on a case that is submitted before it for determination, the court has a duty to juxtapose the facts in the prior decision with those of the present case. Distinguishing is a device used by courts when they are not inclined to follow a particular precedent. Distinguishing cases means pointing out the essential difference between a present case and a prior one, so as not to follow the prior one. Such differences are mostly of facts and not law, though sometimes, the law applicable might be the distinguishing factor (Sanni, 2006: p. 186).
5. Species of Judicial Precedent
Judicial precedent is generally classified into two: “binding precedent” and “persuasive precedent”. A precedent is binding, when a court is obliged to abide by the decision made in an earlier case which is of similar facts as the one before it. On the other hand, a precedent is persuasive, when a court may, but is not obliged to abide by the decision made in an earlier case, even when it is of similar fact with the one before it. However, the effect of a legal precedent on a court i.e. whether of binding or persuasive effect, depends on the level of the court from which the precedent derives. This means, judges of lower court must follow decisions of courts superior to them, but are at liberty to choose whether to follow or not to follow decisions of courts inferior to them or of coordinate level with them (Bello, 2015: p. 65). Conversely, judges of superior courts are not bound to follow decisions of court inferior to them but may be persuaded by them. Also, courts in Nigeria are not bound to follow decisions of foreign courts, they are generally treated as “persuasive authority”, notwithstanding the fact that judges sometimes make reference to foreign judgments in their decisions.
6. Merits of Judicial Precedent
Von Moschzisker asserts that judicial precedent expedites the work of the judges by preventing the constant reconsideration of settled questions, enables lawyers to advise clients with a reasonable degree of certainty and safety, assures individuals that if they act on authoritative rules of conduct their contracts and other interests will be protected in the courts, makes for equality of treatment of all men before the law, and lends stability to the judicial arm of government (Von Moschzisker, 1924: p. 409). He summarises the advantages of stare decisis as certainty, consistency, stability, equality, and predictability (Von Moschzisker, 1924: p. 409).
7. Working Judicial Precedent
The doctrine of judicial precedent has been identified to operate in two ways (Vijayalakshmi, 2011: p. 125), namely:
1) Vertical: this means that prior decisions of higher court are binding on the lower courts; and
2) Horizontal: this means that same court is bound to follow its own prior decision and prior decision of a court of the same level whether past or present.
The practice of judicial precedent is applicable in Nigerian courts and this position has been confirmed in the case of Ojokolobo v. Alamu13, where the Supreme Court held that the doctrine of judicial precedent is a fundamental principle of the Nigerian legal system, and that lower courts are bound by the decisions of higher courts. Nigerian courts adopt both the vertical and the horizontal operation of judicial precedent. By vertical operation, decisions of the superior court (i.e. Supreme Court) bind all courts subordinate to it in the courts hierarchy system14. All courts in the hierarchy system must follow prior decisions of courts higher than itself, and it cannot decline to follow such decisions of a higher court on the ground that it is wrong, otiose or delivered per incuriam (Bello, 2015: p. 69). Legal precedent hangs on the element of superiority (Okeke, 2011: p. 111). In this regard, the decisions of courts of first instance cannot create precedent, because they are not final and are appealable. It is therefore, the decisions of the Supreme Court, as a court of policy, that establishes or creates legal precedents in case law, in Nigeria.
In Nigeria, however, it is observed that, only “persuasive” horizontal precedent is popularly operated among superior courts (Okeke, 2011: p. 72). For example, the Supreme Court is not absolutely bound by its own prior decisions and may overrule itself if the prior decision is found to have been decided in error15 or per incuriam16, Other grounds upon which the Supreme Court will not follow its prior decisions, as garnered from decided cases includes, where there is likelihood of injustice being perpetuated17, and where it would be in the interest of justice and the proper development of the law to do so18. Ditto, the Court of Appeal is, generally, at liberty to decide which of its prior conflicting decisions it will follow. However, it will not follow a prior decision of its own which is arrived per incuriam, or which the Supreme Court has overruled or is likely to overrule.
In the US, the opinions of an appellate court in a State, involving state law, or an appellate court of the United States, a Federal Court of Appeal, or the US Supreme Court involving federal law constitutes a binding precedent on the lower courts19, if such opinion is rendered by a majority of the respective court (Cole, 2013: p. 139). Thus, where it is not decided by a majority of the forum is it rendered as merely a persuasive authority (Cole, 2013: p. 146).
Nonetheless, the US Supreme Court can change its judicial philosophy by overruling its previous decision. Their decision can change as a result of passage of time, evolving culture which renders the precedent obsolete or the basic law may change relating to the precedent by the law maker (Cole, 2013: p. 149). This was the situation in Agostini v. Felton (1997) 521 U.S. 203 where the US Supreme Court changed its judicial philosophy from its earlier decision in Aguilar v. Felton (1985) 473 U.S. 402. In the earlier case, the court had adopted the lemon test in establishing the precedent that, the Establishment Clause prohibited the City of New York from sending public school teachers into church-related schools to provide remedial education to disadvantaged children pursuant to funding for such from a congressionally mandated program (Cole, 2013: p. 148). In Agostini’s case, in departing from the earlier decision, held that “a federally funded program providing supplemental remedial instruction to disadvantaged children on a neutral basis is not invalid under the Establishment Clause, when such instruction is given on the premises of sectarian schools by government employees pursuant to a program containing safeguards such as those present in here.” (Cole, 2013: p. 149).
8. Upturning Judicial Precedent
The strict application of the doctrine of judicial precedent occasions rigidity in judicial determination of cases; as Judges must slavishly kowtow to established precedents on issues submitted before them for determination. Because Judges do not enjoy the latitude to exercise their discretion when presented with cases of established precedents, they are sometimes compelled to give effect to precedents that are antiquated and unreflective of the societal needs and realities. In this section, we shall examine in extenso some of the justification for overturning legal precedents by superior courts in Nigeria and the United States of America.
In Nigeria, the Supreme Court, being the highest Court in the Judicature has the jurisdiction to overturn existing precedents set by it or other lower courts of record in Nigeria20. When a decision of the Supreme Court is made per incuraim, egregiously wrong or arrived at in ignorance of or in concealment of an authority or statutory provision, or likely to perpetuate injustice, the Supreme Court will not be bound in such situation by its previous decision. It can, for such reasons, overrule itself (Anyebe, 2019). In the case of Adegoke Motors Ltd v. Dr. Babatunde Adesanya & Anor (1989) 5 SC 113 at p. 129, Oputa JSC held that:
…it is not part of the jurisdiction or duties of this Court to go on looking for imaginary conflicts. We are final not because we are infallible; rather we are infallible because we are final. Justices of this Court are human-beings, capable of erring. It will certainly be short sighted arrogance not to accept this obvious truth. It is true that this Court can do inestimable good through wise decisions. Similarly, the Court can do incalculable harm through its mistakes… This Court has the power to over-rule itself (and has done so in the past) for it gladly accepts that it is far better to admit an error than to persevere in error…”
This is also the case in the United State of America, as the US Supreme Court has the authority to overturn its prior decisions21. However, the decision to overturn prior precedents must be based on special justification or at least strong grounds that extends beyond the Court’s mere disagreement with the merits of the prior decision’s reasoning (Congressional Research Service, 2018: p. 11). There are several factors that can influence the judge’s desire to overturn an earlier precedent or abide by them. For example, the decision to overturn a judicial precedent can be influenced by the philosophical inclination of the judge concerned (e.g. whether progressive, conservative, or pro-life) or the ideological views or agenda of the judicial officer. We are, however, of the view that overturning of a judicial precedent should not be done without due consideration to the facts and compelling circumstances in the case.
Some of the factors the court must consider in overruling an earlier precedent includes: the quality or soundness of the reasoning in the precedent, workability and/or functionality of the precedent, inconsistency with related decisions, whether other precedents in the area have shifted, changed understanding of relevant facts warranting the precedents (e.g. the US Supreme Court case of Roe v. Wade (1973) 410 U.S. 113) and where the reliance (whether on the earlier decision, even though flawed, and will injure individuals, companies, organizations and the society as a whole who had relied on such decision to acquire legal or other rights (Congressional Research Service, 2018: pp. 12-21)22.
There are two types of overturning judicial precedents, and they are:
1) Direct/explicit overturning: this is when a court explicitly abolishes a rule established by precedents in its jurisdiction and replaces it with the opposite rule (Eisenberg, 2022).
2) Overturning by implication/implicit overturning: this is when a court, instead of explicitly overturning a rule, implicitly overturns it by the process of transformation; in which the court purports to follow a rule while actually undoing it (Eisenberg, 2022). The case of Akeredolu v. Abraham (2018) LPELR-44067 (SC) is illustrative on this, wherein the Supreme Court implicitly overruled the case of Owners of the MV Arabella v. Nigeria Agricultural Insurance Corporation (2008) 11 NWLR (Pt. 1097) 182 on the question of what constitutes “outside jurisdiction” in relation to the Admiralty Court (Federal High Court) for the purpose of determining whether leave of court is required to effect service of an originating process (Erim, 2018).
Notable Cases Where Earlier Precedents Were Overturned in the Jurisdictions Under Consideration
1) The case of Erie R.R. v. Tompkins (1938) 304, US. 64 overruled the case of Swift v. Tyson (1842) 41, US. (16 Pet.) I wherein the Court disapproved the doctrine in Swift holding unconstitutional section 34 of the Federal Judiciary Act of 1789 or any other Act of the Congress.
2) The case of Dobbs v. Jackson Women’s Health Organization (2022) 597 U.S. 215 overruled the case of Roe v. Wade23 and held that the Constitution of the United States does not confer a right to abortion.
3) The case of Yakubu Ibrahim & Ors. v. Simon Obaje (2018) All FWLR (Pt. 937) 1682 which partly overruled the case of Savannah Bank v. Ajilo (1989) 1 NWLR (Pt. 97) 305 and held that the provision for Governor's consent for alienation of interest in land under the Land Use Act, 1978, Cap. 202, LFN 1990 does not apply to land not covered by statutory rights of occupancy, where the alienation is between private individuals, and there is no overriding public interest or conflict between the parties (Ilegieuno et al., 2021).
4) The case of Bucknor-Maclean v. Inlaks Ltd (1980) 8 - 11, SC 1 overruled the cases of Shell B.P v. Jammal Engineering, (1974) 1 ALL N.L.R 543 and Owumi v. PZ (1974) 1 ALL N.L.R Pt. 2 on ground that it had the likelihood of perpetuating injustice.
9. Brief History of AI
The idea of machine being able to function independently of human intervention is considered to be of great antiquity (Tableau). However, for the purposes of this work, we shall limit our consideration to begin from the 20th Century, when AI began to gain wide interest as a field of research to the 21st Century where it has become a mainstream idea
The period between 1950 and 1956 marked the birth of AI. This is the period when the interest in AI became (Tableau). In 1950, Alan Turing published his work “Computer Machinery and Intelligence”. The work birthed what is known today as the “The Turing Test”, which experts use to measure computer intelligence. Also, in 1956, John McCarthy, who is credited as the “father of AI” coined and popularised the term “artificial intelligence” (Teneo.AI). He defined it as “the science and engineering of making intelligent machines.” (Teneo.AI)
The period between 1957 and 1979 is referred to as the “AI Maturation Period” (Tableau). From creation of programming languages, to writing of literatures and films that explored the idea of robots, this period witnessed rapid growth in both research for AI and creation of AI systems (Tableau)24.
The 1980s experienced rapid growth and interest in AI, and has been labelled as the period of “AI boom” (Tableau). By 1985, the AI industry was reported to be more than $1 billion (Singh, 2019: p. 567). At the same time, Japan’s fifth generation computer project stimulated the U.S. and British governments to support academic research in the field (Singh, 2019: p. 567).
Between 1987 and 1993, the AI domain experienced what has been labelled as the “AI Winter”. This period was marked by decrease in private, and government investment in AI, which, in return led to reduced breakthroughs in that period (JavaTPoint, 1955). Despite the draught in funding during the AI Winter, the period between 1993 and 201125 experienced renaissance in funding and support for AI research. During this era, AI professionals shifted their emphasis from attempting to match human intelligence to crafting pragmatic and ingenious software tailored to specific tasks. For example, the year 2002 marked the debut of AI in homes, in the form of Roomba, a vacuum cleaner (Tableau).
From 2012 to the present moment, significant advancements have unfolded within the AI domain (Tableau). There is surge in common-use AI tools, such as virtual assistants, search engines, language translators and lyric creators etc. AI has also found wide applications in different fields of life, including agriculture, education, banking, medicine, space exploration and law etc. These achievements have been attributed to the malgamation of Deep Learning26, extensive data application, and the ongoing quest for artificial general intelligence (AGI)27.
10. Conceptual Clarification of AI
AI (which are Artificial, i.e., manmade, and Intelligence, i.e., thinking power) is a branch of computer science which deals with the study and design of intelligent agents that perceives its environment and takes actions (Singh et al., 2013: p. 1) accordingly to inputs (Ghosh & Thirugnanam, 2021: p. 23). It is characterised as computer programs designed to solve complex problems by implementations of processes similar to human cognition (Singh, 2019: p. 566). It has also been described as the method of simulation of human intelligence designed or programmed by humans (Ghosh & Thirugnanam, 2021: p. 24). Ghosh & Thirugnanam (2021), concludes that, if a system can have the basic skills like learning, reasoning, self-improvement (by learning from experience), language understanding, and solving problems, then it can be assumed that there is the existence of AI.
The theory that “machine will do and think like humans more in the future” is alluded as the concept behind AI (Singh et al., 2013: p. 1). It is well-known fact that humans are considered as the most intelligent and smart species on earth; with the ability to think, apply logic, do reasoning, understand the complexity, and make decisions on their own. Therefore, researchers are trying to create a “man-made homo sapien” species, which can be related to the world of computers in the form of AI (Ghosh & Thirugnanam, 2021: p. 24).
There are many areas which contribute to artificial intelligence which includes mathematics (used for developing algorithms), biology, philosophy, psychology, neuroscience (for studying human mind and its behavior), statistics (for handling huge data), and last but not the least, computer science (to run the algorithm for implementing the concepts). The basic aim of AI is to provide more transparent, interpretable, and explainable systems which can help to establish a better-equipped system used as an intelligent agent (Ghosh & Thirugnanam, 2021: p. 24).
The various types of AI are classified based on their capabilities and their functionalities (Ghosh & Thirugnanam, 2021: p. 24). Based on capability we have: Weak or Narrow AI28, General AI29, and Strong AI30. Based on functionality, we have: Reactive machines31, Limited memory32, Theory of mind33, and Self-awareness34.
11. Ownership and AI: Who Owns Properties35 Generated by AI?
The interplay between intellectual property rights and AI is an important area of intellectual property (IP) law that has gained notoriety with the surging number in AI activity. This has posed a serious difficulty in ascribing ownership of intellectual property in AI-generated works (Ajunwo-Choko & Adeoye, 2024). This is especially so as AI becomes more autonomous making the application of extant IP laws to AI more challenging (Brown, 2020). For example, the recognition of “human authorship” by legal regimes as a parameter for copyright ownership36 prods concern on the ascription of copyright ownership in works independently created by AI. These concerns become more pertinent considering the fact that programming software is considered as proper subject matter of copyright ownership in some jurisdictions (Ezema & Ibekwe, 2022: p. 110).
In Nigeria, under the extant legal regime, creative works generated by AI cannot be meaningfully owned by a human37. Similarly, there is no provision for AI or machine ownership under the extant law. The current position of the law therefore leaves the issue of property ownership in AI in limbo, and creates a vent for endless academic arguments on it. It has been argued that the Copyright Act of 2022, which is the extant law on intellectual property rights in Nigeria, only recognise works with original character (Ajunwo-Choko & Adeoye, 2024). Hence, works created solely by AI, lacking human involvement, do not fulfill the criteria of human authorship and are presently ineligible for copyright protection under the Nigerian Copyright Law (Ajunwo-Choko & Adeoye, 2024). Ezema & Ibekwe (2022), have however argued that, there is always an element of human input in every creativity process of AI which, no matter how minute, often predicts the end-result of the activity. They opine that where a work is created by a machine in Nigeria, the copyright in that work will be conferred on the human inventor or programmer.
The position in the U.S. is analogous to the one in Nigeria. The criterion for ownership of works created by AI is whether there was human creative direction behind the process (Padmanabhan & Wadsworth, 2023: p. 163). Under this statute, the U.S. Copyright Office defines “authorship” to mean “work that was created by a human being.”38 The Office will not register “works produced by a machine or mere mechanical process that operates randomly or automatically without any creative input or intervention from a human author.”39
In Europe, it appears that the legal stance on copyright ownership of Generative AI outputs is still rather unclear and a moot topic. This stems from the backdrop of drought in legislation directly touching on the issue. For example, the European Union (EU) copyright law and recently passed AI Act did not directly address the issue of ownership of AI-generated works (Finkel, 2024; Erickson, 2024). The decision of the Court of Justice of the European Union (“CJEU”) in Infopaq International A/S v. Danske Dagblades Forening40, where it held that, copyright will only be found if there is originality flowing from the author’s own intellectual creation, does provide limited guidance on the issue. By this decision, it appears that the key issue in recognising ownership of AI-generated works is whether there is significant human authorship in the creation process. Similar to the decision in Infopaq is the one laid down by a Czech municipal court in Prague in S. Š. v. Taubel Legal, advokátní kancelář s.r.o., that, under Czech copyright law, a work must be the unique result of the author’s creative intellectual activity and the author can only be an individual. For this reason, an image generated by AI model does not qualify for copyright protection under Czech copyright laws (Chloupek, 2024).
In Germany, their extant laws do not recognise machine authorship of creative works for the purpose of copyright ownership (Finkel, 2024). Likewise in France, the current presumption is that only natural persons can be authors, and originality requires the personal touch or intellectual effort of the author, whereas implementation of automatic and constraining logic without genuine personal effort will not qualify for copyright protection (Finkel, 2024).
Debates are however beginning to heat up in support of the proposition that “legal personhood” should be ascribed to AI to claim ownership of creative works generated by it (Brown, 2020). Amongst the arguments that has been canvassed in support of the proposition are the growing autonomy of AI41 and the need for legal certainty (Chopra & White, 2004). Proponents of this proposition classify AI into weak AI and strong AI (Brown, 2020). Weak AI, which relies on its human programmers and human data input, should own property and enforce its rights and protect its interests through its human programmer or owner. While strong AI, which is able to act independently of human intervention should not be legally allowed to own property (Brown, 2020).
Extant IP regimes are visibly inadequate to address the challenges and opportunities that AI presents. A revamp of the existing IP regimes becomes imperative, to clarify the intricacies surrounding ownership of AI-generated works.
12. Why AI?
AI has become increasingly important in modern societies for a variety of reasons, including (Chojnowska, 2023):
1) Efficiency: AI can automate repetitive and mundane tasks, allowing human user to focus on more complex and creative tasks.
2) Improved decision-making: AI can analyse large amounts of data and make predictions or recommendations based on that data, which can help the human user make more informed decisions.
13. What Role Can AI Play in the Nigerian Justice
Administration System?
The Nigerian judicial system faces significant challenges, which includes delay in justice delivery, lack of transparency, paper-based administrative procedures (Bello & Ogufere, 2024: p. 12), and staggering backlog of cases. The daily influx of cases across all levels of the court system contributes to these challenges. Recently, the erstwhile Chief Justice of Nigeria, Justice Kayode Ariwoola, bemoaned how the Supreme Court is grappling with 6,884 delayed cases, with a workforce of only 13 Justices out of the 21 that is constitutionally provided (Azu, 2022). Comparatively, India Supreme Court, with 34 Justices, has over 80,000 pending cases before it (Rebecca, 2024). This data points to the fact that delay in administration of justice is not necessarily a function of the number of available judges, but other issues such as administrative inefficiency and inadequate use of technology are the other unsung contributory factors. In this context, AI emerges as a powerful tool that can revolutionise and bring about a more efficient and trusted judiciary in Nigeria. By harnessing the capabilities of AI, we can address critical pain points and enhance the efficiency of our courts (Verma, 2024).
The judiciary has been likened to the “text processing industries”, with language, documents, and texts being the raw materials of legal and judicial work (Schindler, 2024). One area in which AI can therefore be most effective to the judiciary is in helping judges and judicial assistants cope with the marmot number of documents that are filed in court on a daily basis. Unfortunately, most of the judicial and clerical works which are repetitive are done manually. By leveraging on AI algorithms, judges can manage their workload more effectively by prioritising cases based on justice demand or other relevant factors, expediting the decision-making process and reducing delays. Administrative tasks can also be streamlined using AI tools, and freeing up judicial resources to handle more complex legal matters (Ogoniba, 2023: p. 63). In Germany, for example, an AI tool called OLGA42 has been reported to potentially reduce the processing time of cases by 50% (Schindler, 2024). With OLGA, judges and judicial assistants can sift through thousands of documents faster and use specific search criteria to find relevant information from various documents (Schindler, 2024).
With AI, administrative processes like case management and allocation can also be optimised. In the Lagos State High Court division for example, it takes a standard 2 to 3 months after filing before cases get allocated to judges. Despite its obvious benefits, majority of Nigerian courts lack a transcription machine and a stenographer to work with. The reality is that judges are beleaguered to capture proceedings in their court in longhand. It is common for a judge to have more than 20 cases on his “cause list” for a day, but may not attend to more than half of the cases due to fatigue from repeated tasks. All these and more culminate to delay in justice delivery by the judiciary. By incorporating AI systems for case administration and resource allocation, such as algorithms that can scan and structure large volumes of evidentiary files for relevant information, or allocate cases to competent court; judges are enabled to concentrate more on substantive case analysis (Smuha & Hendrickx, 2023) without having to waste time on auxiliary and repetitive tasks.
14. The Nexus between AI and Judicial Precedent in Judicial Decision-Making
When judges are presented with legal issues to decide on, they basically not only look for the relevant black letter law, but consider what judges have decided in previous similar cases, i.e., judicial precedent as their source of inspiration and authority. Retrieving, identifying, and analysing the applicable precedent is however a challenging task that the human judges have to confront (Smuha & Hendrickx, 2023). Shouldn’t we then, integrate AI to the judicial decision-making process to assist judges in navigating through these challenges, while they still retain their autonomy to make reasoned judgment?
At this point, it is important to clarify that we are not projecting the idea that AI should take away human autonomy in judicial decision-making. Rather that, by utilising AI tools in the judicial decision-making process, judges can seamlessly assess, retrieve, and analyse data (judicial precedents) which can in turn improve the judges’ understanding of the legal issues presented to them, and remove indiscretion in their exercise of judicial discretion. Think of an algorithm with the capacity to analyse and make distinction between principles laid down in decided cases, derive insights on the socio-economic impact of the cases, and make recommendations. This capacity is especially useful in complex cases where there may be an overwhelming amount of related precedent to be evaluated. The process will afford a Judge who is presented with legal issues to be in vantage position to decide between which authority he will follow, and which one he will not follow, having regard to the facts and issues submitted before him. Although there are certain AI tools already in use in Nigerian courts, which can help judges in retrieving data and information, like the Primsol GPT and PrimeGPT (Giwa & Kodjovi, 2023) however, in terms of their analytical capacity there is need for improvement.
Notwithstanding the potential benefits that have been attributed to the digitalisation of the judiciary, it is also important to acknowledge and address the attendant risks and challenges that the utilisation of AI in the judicial decision-making process portends. One of such risks is the issue of data security. There are concerns that the use of AI tools in legal research may pose high risk to the fundamental rights of individuals (Wiewiórowski & Fila, 2022)43. Another issue of concern is the safety of the data used in tooling AI tools that are utilised by the judicial decision-making process. “Data” is the lifeblood of AI systems, and as Wiewiórowski and Fila noted, the better-quality of data used, the better they are trained (Wiewiórowski & Fila, 2022). Since the majority of systems used in the administration of justice are developed and used for administrative purposes (Smuha & Hendrickx, 2023), it appears less likely therefore, that personal data will be used by programmers in developing AI systems for the use by the judiciary. If it becomes necessary for AI systems to be trained with personal data, the fundamental principle of “data minimization”44 must be followed by programmers. The principle of data minimisation is recognised and adopted under the EU AI Act45.
Another significant concern with the integration of AI to the judicial decision-making process is the issue of “algorithm bias”46 through coding. As Rt. Hon. Sir Robert Buckland expressed, algorithm bias is the most serious weakness of AI’s application to judicial decision-making. It has been argued that the assumptions and biases of the programmers are built into the search algorithms that humans write, and the biases of past researchers are likewise incorporated into the user data that algorithms rely on (John & Nicholas, 2022: pp. 6-7). And because of this, what a researcher finds in the process of searching depends heavily on who programmed the search algorithm and what choices the programmer made in the process (John & Nicholas, 2022: pp. 6-7).
The concern of bias is not unassociated with humans. As Sunstein notes, “the judgments of human beings can be biased; they can also be noisy.” (Sustein, 2022). This is more relevant in light of factors like, the socio-economic background, upbringing, and education of human judges, which shapes their worldview (Buckland, 2023). Indeed, algorithm could help human judges identify and attenuate the cognitive biases that sway their decision-making by improving their accuracy (Sustein, 2022). It is our view that the utilisation of AI in the judicial decision-making process should be of a limited and supporting role only. More like AI being a judicial aide that would assist human judges to assess, retrieve and analyse legal precedents that are applicable to the set of facts or legal issues placed before a judge, to be able to exercise his discretion properly. Even where AI tools assist human judges, the human judges must, as a matter of necessity remain at the epicenter of the decision making.
15. Our Polemics
The interplay of AI and the judiciary has come to stay. As the field of AI develops, it will continue to have a transformative effect on the administration of justice. It seems settled that AI is capable of playing active role in the judicial decision-making process without undermining human autonomy. Judges can leverage on algorithms that are purpose-built for judicial precedent analytics to make reasoned judgments; on when to follow, distinguish, or overturn precedents. Although the need for the application of AI tools in judicial decision-making process is encouraged and has become widely imperative, it is submitted that caution and organic growth in its wholesome application should be intentionally applied. The application of AI in the judicial system should complement human expertise rather than replace it.
As Lord Denning rightly demonstrated throughout his career on the Bench; Judges must not become pedantic slaves to stare decisis. Just as the propositions of the scientist fall to be modified when shown not to fit all instances, or even discarded when shown to be in error, so also precedents should be modified when found to be incongruous to modern time or discarded when found to work injustice (Denning, 1979: p. 292). It is the writers’ projection that by the time AI tools with the capacity to analyse precedents becomes accessible to our judges, a host of existing precedents that have become antiquated and unreflective of the socio-economic dynamics of the country, such as the decision in Okafor v. Nweke (2007)10 NWLR (Pt. 1043) 52147, will be overturned and replaced with more progressive precedents.
Cases
Adegoke Motors Ltd v. Dr. Babatunde Adesanya & Anor (1989) 5 SC 113.
Agbai v Samuel Okogbue (1991) 1 NWLR (Pt. 204) 391.
Agostini v. Felton (1997) 521 U.S. 203.
Aguilar v. Felton (1985) 473 U.S. 402.
Akeredolu v. Abraham (2018) LPELR-44067 (SC).
Apple Computer, Inc. v. Formula Int’l, Inc., 725 F.2d521 (9th Cir. 1984)
Apple Computer, Inc. v. Franklin Computer Corp., 714 F.2d 1240 (3d Cir.1983)
Aqua Limited v. Ondo State Sports Council (1988) JELR-42826 (SC).
Bashire M. Dalhatu v. Ibrahim S. Turaki (2003) 15 NWLR (Pt. 843) 310.
Bucknor-Maclean v. Inlaks Ltd (1980) 8-11 S.C. 1.
Chevron USA Inc. v. Natural Resources Defense Council, Inc. (1984) 467 U.S 837.
Dairo v. UBA Plc (2007) 16 NWLR (Pt. 1059) 159.
Dobbs v. Jackson Women’s Health Organization (2022) 597 U.S. 215.
Eliochin (Nig.) Ltd v. Mbadiwe (1986) 1 NWLR (Pt. 14) P. 47
Emeka v. Okadigbo (2012) 18 NWLR (Pt. 1331) 96 (SC).
Erie R.R. v. Tompkins (1938) 304, US. 64.
Infopaq International A/S v. Danske Dagblades Forening (Case C-5/08)
Johnson v. Lawanson (1971) 7 NSCC 82.
Loper Bright Enterprises v. Ramondo (2024) 144 S.Ct. 2244.
Minersville School District v. Gobitis (1940) 310 U.S. 586.
Network Securities Limited v. Dahiru (2022) 14 NWLR (Pt. 1850) P. 351.
Odi v. Osafile (1985) SC.144/1983.
Ojokolobo v. Alamu (1987) 3 NWLR (Pt. 61) 377.
Okafor v. Nweke (2007) 10 NWLR (Pt. 1043) 521.
Okonji Ngwo & Ors. V. Raphael Monye & Ors. (1970) LLJR-SC.
Owners of the MV Arabella v. Nigeria Agricultural Insurance Corporation (2008) 11 NWLR (Pt. 1097) 182.
Owumi v. PZ (1974) 1 ALL N.L.R Pt. 2.
Roe v. Wade (1973) 410 U.S. 113
Savannah Bank v. Ajilo (1989) 1 NWLR (Pt. 97) 305.
Shell B.P v. Jammal Engineering, (1974) 1 ALL N.L.R 543.
Swift v. Tyson (1842) 41, US. (16 Pet.).
West Virgina State Board of Education v. Barnette (1943) 319 U.S. 624.
Yakubu Ibrahim & Ors. v. Simon Obaje (2018) All FWLR (Pt. 937) 1682.
Legislations
Constitution of the Federal Republic of Nigeria (as amended) 2023.
European Union Artificial Intelligence Act, 2024.
Land Use Act, 1978, Cap. 202, LFN 1990.
Legal Practitioners Act, Cap. 207, LFN 1990.
Nigeria Copyright Act, 2022.
The Supreme Court Act, 1960.
Supreme Court (Additional Original Jurisdiction) Act of 2002, Cap. S16 LFN 2004.
U.S. Congress. (1958) United States Code: Copyright Office, 17 U.S.C. §§ 201-216.
NOTES
1This principle is the foundation of the Prediction Theory of Law. For more elucidation on the foundations of this theory, see, Iheme (2024). Defects of English Rules of Contractual Interpretation and their Challenges for African Businesses. AJLS, pp. 351 379.
2Throughout this paper, the terms “judicial precedent” and “legal precedent” will be used interchangeably.
3See the case of Eliochin (Nig.) Ltd v. Mbadiwe (1986) 1 NWLR (Pt. 14) P. 47, where the Supreme Court held that: “since the abolition of the Privy Council from the hierarchy of our Courts and the enactment of the Nigerian Republican Constitution in 1963, English Courts’ decisions are no more binding on Nigerian Courts but have persuasive, even if essential value.”
4For example, the United Kingdom practices a constitutional monarchy and parliamentary democracy system of government while, in Nigeria, a federate and parliamentary political structure is observed.
5Although, a UK Artificial Intelligence Bill appears to be on the card from the new Labour Government of the UK. Likewise in Nigeria, the Federal Ministry of Communications, Innovation and Digital Economy (FMCIDE) and the National Information Technology Development Agency (NITDA) have reportedly completed the first draft of the National Artificial Intelligence Policy (NAIP) which is set to craft a framework for the overall regulation of AI in the country.
6Or, in its complete form, stare decisis et non quieta movere is usually translated “to stand by (or adhere to) decisions and not to disturb what is settled.”
7Agbai v Samuel Okogbue (1991) 1 NWLR (Pt. 204) 391.
8This is the process whereby the obiter dictum in an earlier case is endorsed by a superior court, in a latter case.
9(2007) 16 NWLR (Pt. 1059) 159.
10See, Bashire M. Dalhatu v. Ibrahim S. Turaki (2003) 15 NWLR (Pt. 843) 310.
11See, Dairo v. UBA Plc (supra).
12See, Emeka v. Okadigbo (2012) 18 NWLR (Pt. 1331) 96 (SC).
13(1987) 3 NWLR (Pt. 61) 377.
14The hierarchy of courts in Nigeria is stratified in the following order: 1) The Supreme Court; 2) The Court of Appeal; 3) The High Courts (Federal and States); The Sharia Court of Appeal and the Customary Court of Appeal in the Federal Capital Territory and the states; The National Industrial Court and other specialised courts; 4) The Magistrate Court; and 5) Area Courts or District Courts of various grades; Customary and native Courts.
15See, Johnson v. Lawanson (1971) 7 NSCC 82.
16See, Okonji Ngwo & Ors. V. Raphael Monye & Ors. (1970) LLJR-SC.
17See, Odi v. Osafile (1985)-SC.144/1983; Aqua Limited v. Ondo State Sports Council (1988) JELR-42826 (SC).
18See, Bucknor-Maclean v. Inlaks Ltd (1980) 8-11 S.C. 1.
19Lower court referred to above mean either the trial or appellate courts on decisions made by the supreme court.
20The Supreme Court Act, 1960 and Section 6(5) of the Constitution of the Federal Republic of Nigeria, 2023 (as amended).
21See, the case of West Virgina State Board of Education v. Barnette (1943) 319 U.S. 624 where the US Supreme Court overturing its earlier ruling in Minersville School District v. Gobitis (1940) 310 U.S. 586, held that the First Amendment protects students from being compelled to salute the American flag or say the Pledge of Allegiance in public schools. The Court in Loper Bright Enterprises v. Ramondo (2024) 144 S.Ct. 2244 overturned the decision in Chevron USA Inc. v. Natural Resources Defense Council, Inc. (1984) 467 U.S 837.
22See, Okafor v. Nweke (2007)10 NWLR (Pt. 1043) 521.
23(1973) supra.
24For example, John McCarthy created “LISP” (acronym for List Processing), the first programming language for AI research, which is still in popular use to this day.
25This period has been labelled as the “Intelligent Agents” period.
26Deep learning is a method in artificial intelligence (AI) that teaches computers to process data in a way that is inspired by the human brain. Deep learning models can recognise complex patterns in pictures, text, sounds, and other data to produce accurate insights and predictions.
27Artificial general intelligence (AGI) refers to the hypothetical intelligence of a machine that possesses the ability to understand or learn any intellectual task that a human being can.
28It is a type of AI which can perform a predefined narrow set of instructions without exhibiting any thinking capability. It is the most widely used type of AI in this world. Some famous examples are Apples’s Siri, Alexa, Alpha Go, IBM’s Watson supercomputer, Sophia (the humanoid) all belong to the weak AI type.
29It is the type of AI which can perform the tasks like what human can do. Although, till now, that has not been achieved. there are no such machines which works like human or can think as perfectly as human, but it may happen in near future.
30it is the type of AI in which it is expected that the machine will surpass the capacity of human.
31These are the machines which works on the data available in the form of predefined dataset. It does not have the facility of data storage for storing the past and future data. It completely depends on the present data.
32These are the machines which can store the past experience or store the memory for limited period of time. An example for limited memory AI is the self-driving cars (it can store the information like speed, distance, speed limit required for the navigation of the car).
33These are types of machines, which are expected to understand the psychological and emotional aspects of human mind and work accordingly.
34These machines belong to a hypothetical concept that will be considered as super-intelligent machines, which can think, act, and will be self-aware as they will have consciousness and sentiments like humans.
35Throughout this paper, the term “AI-generated properties” will be used interchangeably with “AI-generated works”.
36See, for example, the Nigerian Copyright Act, 2022, and 17 U.S.C. § 102 of the United States of America.
37See, Section 2 (1) of the Nigerian Copyright Act, 2022. The types of works eligible for copyright protection under the Act includes literary works, musical works, artistic works, audiovisual works, sound recordings and broadcasts.
38See, U.S. Copyright Office, Compendium of U.S. Copyright Office Practice § 306 (3rd ed. 2021).
39Ibid. § 313.2.
40(Case C-5/08).
41According to the Committee on Legal Affairs of the European Parliament, the more autonomous robots are, the less they can be considered to be simple tools in the hands of other actors (such as the manufacturer, the operator, the owner, the user, etc.); whereas this, in turn, questions whether the ordinary rules on liability are sufficient or whether it calls for new principles and rules to provide clarity on the legal liability of various actors concerning responsibility for the acts and omissions of robots where the cause cannot be traced back to a specific human actor and whether the acts or omissions of robots which have caused harm could have been avoided. See, European Parliament (EP) (2017) “Motion for a European Parliament Resolution” CLA 2015/2103(INL), 27 January 2017
(https://www.europarl.europa.eu/doceo/document/A-8-2017-0005_EN.html). Accessed 8/8/2024.
42OLGA is an AI assistant created by IBM®.
43AI systems that may be used to assist judicial authorities in factual and legal research, as well as in interpreting and applying the results of such research in a specific case are classified as “high risk” AI systems. See, Art. 6(2) EU AI Act, 2024.
44Data minimization is a privacy principle that limits the collection and storage of personal data to what is necessary to accomplish a specific purpose. It’s a key principle for reducing the risk of data breaches and respecting individual privacy.
45See Recital 69 and Art. 10 EU AI Act.
46Algorithmic bias describes systematic and repeatable errors in a computer system that create “unfair” outcomes, such as “privileging” one category over another in a way different from the intended function of the algorithm.
47In this case, the Supreme Court held that it is only a legal practitioner as defined under Section 24 of the Legal Practitioners Act, Cap 207, LFN 1990 can practice as a barrister or as a barrister and solicitor, either generally or for the purpose of any particular office proceeding. A law firm is not a legal practitioner and therefore cannot practice as such by filing processes in Nigerian court. Only human beings actually called to the Bar can practice or practice by signing document. In view of this principle, the Court declared the motion on notice filed on 19/5/05, the proposed notice of cross-appeal and the applicants’ Brief of Argument in support of the motion as incompetent in that they were not issued by a legal practitioner known to law.
Almost two decades after, the same principle of law is still being applied by the same court in the case of Network Securities Limited v. Dahiru (2022) 14 NWLR (Pt. 1850) P. 351. However, the dissenting opinion of Agim JSC in this case is notable, wherein he stated that: “I think that we should depart from Okafor v. Nweke (supra) and similar decisions and return to Buhari v. Yabo because grave injustice will continue to result from following the judicial precedent in Okafor v. Nweke and similar cases…”