Explainable Large Language Models & iContracts

Georgios Stathis

Institute of Tax Law and Economics, Leiden University, Kamerlingh Onnes Building,

Keywords:

LegalTech, Intelligent Contracts, Large Language Models, Explainable AI, User Trustworthiness.

Abstract:

Contract automation is a ﬁeld of LegalTech under Artiﬁcial Intelligence (AI) and Law that is currently under-

going a transition from Smart to Intelligent Contracts (iContracts). iContracts aim to full contracting automa-

tion. Their main challenge is ﬁnding a convincing direction for market adoption. Two powerful market factors

are the advent of Large Language Models (LLMs) and AI Regulation. The article investigates how the two

factors are able to inﬂuence the market adoption of iContracts. Our Research Question reads: to what extent

is it possible to accelerate the adoption of Intelligent Contracts with Explainable Large Language Models?

Following a literature review our research employs three methodologies: market gap analysis, case study, and

application. The results show a clear way for iContracts to follow, based on existing market gaps. Moreover,

they validate whether the application of Explainable LLMs is possible. The discussion clariﬁes the main limi-

tations with Explainable LLMs. Our conclusion is that the two factors are impactful for so long as the market

adoption attempts to bridge the gap between innovators and early adopters.

1 INTRODUCTION

Smart Contracts have laid the foundation for iCon-

tracts. Smart Contracts are self-executing contracts

with the terms of the agreement directly written into

code (Madir, 2020). They are rooted in blockchain

technology, which provides transparency and trust in

the digital realm (Werbach, 2018).

While Smart Contracts have revolutionised the

contracting process by automating transactions and

reducing the need for intermediaries, they have lim-

itations. Smart Contracts are essentially binary, and

capable of executing predeﬁned actions, but inca-

pable of interpreting and adapting to complex legal

nuances (Mik, 2017).

iContracts represent the next step in the evolu-

tionary process. They are designed to go beyond

the straightforward nature of Smart Contracts. The

key objective is to enable iContracts to handle au-

tonomously the entire contracting process, encom-

passing everything from negotiation to execution.

Such an action path entails a seamless integration of

human-readable language and code (Mason, 2017).

As a result, iContracts will possess the capacity to un-

derstand, adapt, and evolve in response to the intrica-

cies of real-world contracts, thus paving the way for

https://orcid.org/0000-0002-4680-9089

a new era of automation (Stathis et al., 2023c; Stathis

et al., 2023b).

• Deﬁnition 1: An intelligent contract or iCon-

tract is a contract that is fully executable without

human intervention

Meanwhile, the main challenge with iContracts

is the market adoption rate (McNamara and Sepas-

gozar, 2018). Automation is closely related with the

law. Owing to the high legal consequences of iCon-

tracts, human users prefer traditional methods such as

the direct inclusion of a legal expert (Stathis et al.,

2023a). Such a preference is closely connected with

two developing trends: (1) the advent of Large Lan-

guage Models (LLMs) and (2) the regulation of AI.

The adoption of LLMs has happened at the speed of

light. As a result we may wonder how long the adop-

tion of iContracts will take. Is it ﬁve years or only half

a year? Next to this question we are facing (1) the

regulation of AI, with the AI Act being currently in

preparation (European-Parliament, 2023), and (2) the

global battle to regulate technology (Bradford, 2023).

One of the main challenges with AI technologies is

the lack of user trustworthiness (Liang et al., 2022).

Research shows that by understanding how AI takes

https://bravenewcoin.com/insights/pamela-morgan-

at-bitcoin-south-innovating-legal-systems-through-

blockchain-technology

1378

Stathis, G.

Explainable Large Language Models & iContracts.

DOI: 10.5220/0012607400003636

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 16th International Conference on Agents and Artiﬁcial Intelligence (ICAART 2024) - Volume 3, pages 1378-1385

ISBN: 978-989-758-680-4; ISSN: 2184-433X

decisions the outcome of the decision can then be ex-

plained and hence user trustworthiness increases (Xu

et al., 2019).

• Deﬁnition 2: Large language models (LLMs)

are a category of foundation models trained on

immense amounts of data making them capable

of understanding and generating natural language

and other types of content to perform a wide range

of tasks

Our motivation for performing this research is to

improve the rate of market adoption of iContracts. We

believe that by examining to what extent it is possible

to apply Explainable LLMs on iContracts, we are able

to increase user trustworthiness among humans.

The aforementioned thoughts lead us to the fol-

lowing Research Question (RQ).

RQ: To What Extent Is It Possible to Accelerate

the Adoption of Intelligent Contracts With Explain-

able Large Language Models?

The article’s contribution is that it shows and val-

idates whether the application of Explainable LLMs

on iContracts is possible to direct the future attempts

in developing iContracts ﬁt for market adoption.

To answer the RQ, we structured the paper as fol-

lows. In Section 2, we describe the relevant literature.

Section 3 presents the methodology and Section 4 the

results. Section 5 discusses the results and focusses

on theoretical and practical parameters. Finally, Sec-

tion 6 answers the RQ and provides the conclusion

together with a preview on the future.

2 LITERATURE

The literature present sources on: the three contract-

ing revolutions (2.1), market adoption in business

studies (2.2), the application of LLMs in contracts

(2.3) and the development of trustworthy and Ex-

plainable LLMs (2.4).

2.1 Three Revolutions

The path from human contracts to iContracts is char-

acterised by three revolutions. The ﬁrst revolution

started with the transition from physical contracts to

digital contracts (also known as Electronic Contracts

or eContracts) (Krishna and Karlapalem, 2008). With

the advance of big data, people gradually digitised

physical contracts into electronically accessible doc-

uments (von Westphalen, 2017). Moreover, they re-

placed some physical labor such as signing, by elec-

tronic handling. Today, most market developments

https://www.ibm.com/topics/large-language-models

are concentrating on further adopting and expanding

the adoption of digital contracts

In the last decade (2010-2020), with the advent

of blockchain technology the second revolution oc-

curred, which switches the focus from digital con-

tracts to smart contracts. Smart contracts are agree-

ments that are executable by code, most often on

blockchain-based distributed ledgers (Khan et al.,

2021). The promise of smart contracts is to be seen

as the replacement of human language by human

code. However, the adoption of smart contracts, be-

yond specialised blockchain or experimental market

circles, has not managed to reach automatically wide

market adoption

. One of the main challenges of

smart contracts is that it is hard for users to understand

and trust the computer code behind them (Zheng

et al., 2020).

Here we arrive at the third revolution, the transi-

tion from smart contract to iContracts. iContracts aim

at full contracting automation with minimal to no hu-

man involvement (Stathis et al., 2023c; Stathis et al.,

2023b). iContracts promise to bridge the gap that

smart contracts have been unable to ﬁll. ”Bridging”

will gain user trustworthiness by combining computer

code with user-friendly, readable and understandable

code, that originates from physical contracts (McNa-

mara and Sepasgozar, 2020). Despite the surge in

scientiﬁc interest in iContracts for the construction

industry, there remains a signiﬁcant demand for tra-

ditional contracting(ﬁrst recognised in 2020 (McNa-

mara, 2020)); meaning the development of iContracts

is still in its formative stages.

As a follow-up, we mention a research pro-

posal (Stathis et al., 2023c; Stathis et al., 2023b) pre-

sented at International Conference on Agents and Ar-

tiﬁcial Intelligence (ICAART) 2023, which revived

research on iContracts by re-purposing use cases, in-

stead of delegating the task of the complex construc-

tion industry. A reference to a straightforward free-

lancer agreement would simplify the evident com-

plexity in construction contracts. In particular, such

freelance agreements stipulated the extent to which

iContracts may contribute in speciﬁc domains. With

increasing attention to the details of iContracts in sim-

pler domains it is possible to gradually expand into

more complex domains. Notwithstanding these de-

velopments, it is still unclear how to develop a path

towards the market adoption, that contributes towards

end-users increasingly adopting iContracts.

https://www.gartner.com/en/documents/3981321

https://www.grandviewresearch.com/industry-

analysis/smart-contracts-market-report

Explainable Large Language Models & iContracts

1379

2.2 Market Adoption in Business

Studies

In business studies, market adoption has been stud-

ied for many years (Posthumus et al., 2012; Botha,

2019). The most important contribution is Geoffrey

Moore’s Crossing the Chasm. It studies how tech-

nologies penetrate a group ranging from (1) innova-

tors to (2) early adopters and gradually advance into

(3) wide adopters, (4) late adopters until they reach

the so called (5) laggers (Moore, 1991). The re-

sults are general: every technology that enters mar-

ket adoption follows the same route (Goldasteh et al.,

2022). We are going to examine the adoption of

iContracts through the same ﬁve lenses. Each con-

tracting revolution can be located at a speciﬁc stage

in Moore’s Crossing the Chasm. By identifying the

speciﬁc position of iContracts relative to digital and

smart contracts, it is (1) also possible to identify rel-

evant gaps that require to be bridged, and then (2) to

investigate the extent to which LLMs can contribute

in bridging the gaps.

2.3 Large Language Models in Contract

Automation

Since the dissemination of LLMs with ChatGPT, sci-

entists and business people have attempted to apply

LLMs, falling under the category of Generative AI,

on contracts (see for example recent eighty million

fundrasing for Silicon Valley based Harvey

). The

extent and results of such applications are still largely

unclear. On the one hand, end-users expect that gen-

erative AI tools assist them in developing contracts at

a fraction of the costs that it would usually cost to pay

a lawyer

. On the other hand, the generated contracts

(seen as LLMs) are not necessarily perceived as trust-

worthy, as with most outputs of LLMs and Generative

AI (Lenat and Marcus, 2023). They may ﬁnd results

vague, confusing or even mysterious (potentially the

result of hallucination). Hence, due to the high impact

of legal consequences, end-users (even if they do at-

tempt themselves to generate a contract) still prefer to

consult their lawyer (Stathis et al., 2023a). The larger

the degree of the severity of legal consequences, the

lower the reliance on generative contracts (Stathis

et al., 2023a). Some scientists might argue, that in a

way Generative Contracts are announcing the fourth

contracting revolution (Williams, 2024). From our

https://siliconangle.com/2023/12/20/harvey-raises-

80m-build-generative-ai-legal-professionals/

https://www.docusign.com/blog/products/generative-

ai-contracts-agreements

perspective, we consider Generative Contracts to be

only a feature (i.e., one aspect) of iContracts (or of

Smart and Digital contracts) and not the architecture

behind iContracts. In the methodology, result and dis-

cussion sections we will clarify why we believe that

this is so.

2.4 Trustworthy and Explainable Large

Language Models

In 2024, we may expect that the European Union

(EU) will approve the very ﬁrst rules for AI in the

world (European-Parliament, 2023). The rules will be

presented in the AI Act, a regulation which is aiming

to regulate two kinds of AI: Predictive AI and Gener-

ative AI (Stathis and van den Herik, 2023). The main

focus of these rules is to make sure AI can be trusted

in accordance with the directions provided to the EU

by the High-Level Expert Group on AI (HLEG-AI,

2019). Here, we see two sides: (1) society seems to

miss trust towards AI, and (2) the EU wants to protect

its people by developing rules to prevent negative con-

sequences by the wide adoption of AI (Lockey et al.,

2021). This is presisely the reason why some AI tech-

nologies are completely prohibited and other AI tech-

nologies are considered as High-Risk, which are also

subject to most of the regulatory restrictions

. Gen-

erative AI are treated as an exceptional AI technol-

ogy which should adhere to speciﬁc restrictions, due

to the high reliance on trained data which are poten-

tially subject to copyright laws

(Helberger and Di-

akopoulos, 2023). Due to the difﬁculty in decipher-

ing levels of perceived trustworthiness from actual

trustworthiness, scientists emphasise the concepts of

transparency and accountability in matters related to

trustworthy AI (Munn, 2023). The main principal

way along which an AI-system can gain transparency

and accountability is via explainability (Holzinger

et al., 2020). Explainable AI is the ﬁeld in which

researchers investigate how AI system decisions can

be explained in an understandable, interpretable and

trustworthy manner for humans

. In the Springer

AI and Ethics Journal researchers deal with the ques-

tion: how the explainability issue can be handled with

https://www.consilium.europa.eu/en/press/press-

releases/2023/12/09/artiﬁcial-intelligence-act-council-and-

parliament-strike-a-deal-on-the-ﬁrst-worldwide-rules-for-

ai/

https://www.consilium.europa.eu/en/press/press-

releases/2023/12/09/artiﬁcial-intelligence-act-council-and-

parliament-strike-a-deal-on-the-ﬁrst-worldwide-rules-for-

ai/

https://www.marktechpost.com/2023/03/11/under–

standing-explainable-ai-and-interpretable-ai/

ICAART 2024 - 16th International Conference on Agents and Artiﬁcial Intelligence

1380

the development of a ”culture of explainability” that

supports the explicit explanation of the architecture,

design, production, and implementation of decision-

making across the AI development and implementa-

tion chain (Stathis and van den Herik, 2023). The

result is very important, even vital for our research,

provided that for the case of applying Generative AI

on iContracts, such explainability culture will impact

the trustworthiness of the end user signiﬁcantly. As

positive implications we particularly foresee positive

developments for ethical and legal transparency and

accountability.

3 RESEARCH METHODOLOGY

The research methodology starts with an analysis of

the market gap (3.1) and then introduces the case

study (3.2) as well the Generative AI application in

close coherence with LLMs (3.3).

3.1 Market Gap Analysis

Our methodology begins with the visualisation of

the three revolutions into a graph based on Moore’s

Crossing the Chasm graph (Figure 1). This visuali-

sation will allow us to identify the positioning of the

three technologies (digital, smart, and intelligent con-

tracts) in relation to market adoption. Thereafter, we

are able to conduct a gap analysis to identify the steps

that appear to be the missing links for the adoption of

iContracts. Our gap analysis will focus on a speciﬁc

case study.

To classify market categories on contracting au-

tomation we are going to use Legalcomplex’s cate-

gorisation

. Legalcomplex is the largest database

on LegalTech solutions. Legalcomplex has classiﬁed

contract automation solutions into the following ﬁve

categories: (1) contract negotiation, (2) contract risk

management, (3) contract drafting, (4) contract ex-

traction and (5) contract management.

The classiﬁcation of Legalcomplex follows the

typical journey of a contracting user. Starting with

negotiations, a legal expert makes an estimation of

legal risk and drafts a contract. Thereafter, relevant

elements of that contract can be extracted during the

execution stage and/or the monitoring stage; ﬁnally

the contract is being discussed on management level

until its completion.

https://legalcomplex.com

3.2 Case Study

The case study is presented by the Knowledge Graph

developed in the ICAART 2023 research and is based

on the Onassis Ontology (see Figure 2). The Knowl-

edge Graph validates how a small agreement (be-

tween two contractors, guided or supervised by a le-

gal expert) can be automated based on communica-

tions and risk data that are exchanged between two

contracting parties (Figure 2). In the communication

with the contractor, the back-end the role of a legal

expert is to clarify the scope, contract, risks and ques-

tions for the contracting parties. When (1) contracting

parties (2) select a scope, (3) the reply questions and

(4) the risk-intelligent contract is updated by the vari-

ables from the conversation exchange (see in Figure

2 the lines: question, answer, section, contract, vari-

able). The question then is, how can Generative AI

help? Our experiment will guide the reader to an an-

swer to this question.

3.3 Application

The application shows to what extent is it possible

to develop alternative options by Explainable LLMs

to facilitate the automated generation of data for the

stakeholders involved in our case study (i.e., the legal

expert and the contractors). Starting with the KG, we

will ﬁrst identify the potential locations where Gen-

erative AI and LLMs can be applied. Then we are

showing the power of Google’s Bard LLM to validate

whether and to what extent it is indeed possible to

develop explainable generated data that functionally

assist the iContract users into completing their work.

To amplify the explainability capabilities of Genera-

tive AI and LLMs, in compliance with the literature

on trustworthy AI, we will use the opportunity to re-

quest the LLM model to provide higher levels of user

trustworthiness with alternative explanations.

4 RESULTS

The results visualise the market gap analysis (4.1) and

provide the experiment validation (4.2).

4.1 Market Gap

We start with the visualisation of the market adop-

tion of contracting technologies in accordance with

Moore’s Crossing the Chasm (Figure 1) (Moore,

1991). First, we see that physical contracts are the

preferred contracting method for laggards. Then, dig-

ital contracts are reaching the late market. Here,

Explainable Large Language Models & iContracts

1381

Figure 1: Moore’s Crossing The Chasm.

we remark that Smart contracts have not managed to

achieve a wide adoption of crossing the chasm. Third,

at the same time, iContracts have still not reached

early adopters and are therefore being evaluated by

innovators only.

Based on (1) this visualisation and (2) the user

preferences, we are able to identify the relevant mar-

ket gap among each contracting alternative. Using

as comparison framework Legalcomplex’s classiﬁca-

tion of contract automation solutions, we see in Ta-

ble 1 (column 2) that the highest adoption is observed

with physical contracts. The main obstacle for phys-

ical contracts is data extraction. That obstacle has

been solved by digital contracts. Hence, gradually

digital contracts are the ﬁrst followers. Still digital

contracts have not managed to replace negotiations or

risk management as they occur in physical contracts.

Also contract drafting seems to have remained signif-

icantly reliant on physical contract drafting (although

it is nowadays done often by electronic means, even

with the use of templates). Smart contracts (see col-

umn 4) have lower levels of adoption across most cat-

egories, except with drafting, extraction and manage-

ment which is ampliﬁed by electronic means. With

intelligent contracts, we see the lower adoption rate.

The table helps us to decipher which gap iContracts

should aim to bridge ﬁrst. That is the gap in contract

negotiations, risk management and drafting. As seen

in Figure 2, this is the conceptual line of execution

that the Onassis Ontology is taking. Hence, by ex-

perimenting with the application of Explainable Gen-

erative AI and LLMs on the Onassis Ontology, it is

expected to bridge the market gap observed on these

three categories faster.

4.2 Explainable Generative AI

Validation

The application results, which are accessible on

GitHub (

) validate that Generative AI can be lever-

aged in every single step of the Onassis Ontology. For

as long as the Generative AI prompt requires the spec-

iﬁcation of supporting explanations, the provision of

such explanations is possible. As a direct conse-

quence, we see that it becomes visible that the reduc-

tion of human labour in iContracts is further ampli-

ﬁed with the help of Generative AI. By including the

explicit provision of explanations we may also expect

that the user trustworthiness increases. Hence, tech-

nologically, it is possible to support the gap difference

between iContracts with other contracting alternatives

with Explainable Generative AI.

5 DISCUSSION

In our discussion we focus on the implications for the

adoption of iContracts (5.1) and the limitations ob-

served in the Explainable LLM application (5.2).

5.1 iContracts Market Adoption

Implications

As we have seen, the market gap analysis became use-

ful in identifying the precise gap between the alterna-

tive contracting options for end users. We found that

https://github.com/onassisontology/onassisontology/

blob/main/img/EGENAIEXP.png

ICAART 2024 - 16th International Conference on Agents and Artiﬁcial Intelligence

1382

Figure 2: Visualisation of the use case scenario modelled with Onassis’s ontological expressiveness.

the advantage of iContracts lies exactly where digital

and smart contracts seem to struggle with replacing

Physical contracts, namely contract (1) negotiations,

(2) risk management and (3) drafting.

In compliance with market adoption, business

study literature, provided that physical, digital and

smart contracts already present certain alternatives

that users have habituated. The opportunity for iCon-

tracts to impress early adopters and gradually wide

adopters, lies in exploiting opportunities that signiﬁ-

cantly impact current practices.

In conclusion, there is thus an opportunity by ap-

plying Explainable Generative AI, which simpliﬁes

human complexity in multiple directions, as validated

by the application on the case study. Still, such appli-

cation allows us also to identify speciﬁc limitations

which we present below.

Explainable Large Language Models & iContracts

1383

Table 1: Rate of contracting adoption based on high-mid-low factors among four alternative contracting options.

Physical Digital Smart Intelligent

Contract Negotiation High Mid Low Low

Contract Risk Management High Mid Low Low

Contract Drafting High Mid Mid Low

Contract Extraction Low High Mid Low

Contract Management High High Mid Low

5.2 Explainable Large Language Model

Limitations

Requesting an LLM to develop supporting explana-

tions is a really possible challenge. Such explanations

improve user trustworthiness by supporting their eval-

uation process. Instead of requesting an LLM to di-

rectly provide one single answer for a given topic, the

alternative option is to allow for users to select the

most serviceable answer for a speciﬁc purpose.

If we compare the effectivity of humans to pro-

duce alternative explanations against an LLM produc-

ing such alternatives, we ﬁnd that it is a laborious task

that most humans would avoid due to increased com-

plexity. Hence, an Explainable LLM is able to replace

a laborious task with a more trustworthy alternative.

An additional positive aspect is that we validated

the application of Explainable LLM as being possible

across each step of the Onassis Ontology. Indeed, we

found that the application of Generative AI on iCon-

tracts happens on feature level and not on architec-

ture level. Although, our request for explanations still

presents certain limitations, which are not yet solved.

The limitations are related with the inability of

Generative AI systems to provide (1) readable sources

that verify its explanations and (2) understandable

reasoning patterns that explain the reasoning of its de-

velopers. In compliance with the ”culture of explana-

tion” as supported in the Springer research, these two

characteristics will be able to drastically increase the

user trustworthiness even further, due to the ability for

users to validate the origin of speciﬁc data as well as

the line of reasoning a developer has followed. Such

transparency helps with improving end user evalua-

tions as well as with assigning liability with improved

accountability that also increases user trustworthiness

from a legal point of view.

6 CONCLUSION

The RQ of this research is: to what extent is it pos-

sible to accelerate the adoption of Intelligent Con-

tracts with Explainable Large Language Models?.

The answer is that Explainable LLMs, as a cat-

egory of Generative AI, have the possibility to ac-

celerate the adoption of iContracts to the extent that

it is applies on the categories that are currently un-

derrepresented by the competing contracting alterna-

tives, namely contract (1) negotiations, (2) risk man-

agement and (3) drafting. This occurs because of two

characteristics for Explainable LLMs. First, Gen-

erative AI can make the laborious human tasks in-

volved in the three categories simpler. Second, the

explainability aspect can increase the end user trust-

worthiness signiﬁcantly by supporting the outputs of

Generative AI with explicit explanations. To further

leverage explainability, we advice for being compliant

with the ”culture of explanations”. Hence, the devel-

opers of LLMs should be connaisseurs of the culture

in which the system operates when given the task to

make an explicit presentation of sources supporting

data outputs as well as explicit representation of the

line of reasoning supporting algorithmic models.

Our further research will focus on developing ex-

periments with additional case studies, with gradually

increasing complexity. Owing to a high reliance on

end user reactions, we will continue to conduct end

user validation experiments with graphical user in-

terfaces.

The overall novelties of the paper is that (1) it

presents the state of market adoption of contracting

technologies, (2) it identiﬁes the gap among the alter-

native contracting technologies, (3) it connects the as-

pect of explainability with LLMs, and (4) it concludes

that the application of Explainable LLMs is success-

fully possible on practical case studies.

REFERENCES

Botha, A. P. (2019). Innovating for market adoption in the

fourth industrial revolution. South African Journal of

Industrial Engineering, 30(3):187–198.

Bradford, A. (2023). Digital Empires: The global battle to

regulate technology. Oxford University Press.

European-Parliament (2023). Artiﬁcial Intelligenc Act:

Proposal for a regulation of the European Parliament

and of the Council laying down harmonised rules on

ICAART 2024 - 16th International Conference on Agents and Artiﬁcial Intelligence

1384

artiﬁcial intelligence (artiﬁcial intelligence act) and

amending certain Union legislative acts. Eur Lex, Eu-

ropean Union.

Goldasteh, P., Akbari, M., Bagheri, A., and Mobini, A.

(2022). How high-tech start-ups learn to cross the

market chasm? Journal of Global Entrepreneurship

Research, 12(1):157–173.

Helberger, N. and Diakopoulos, N. (2023). Chatgpt and the

AI Act. Internet Policy Review, 12(1).

HLEG-AI (2019). High-level expert group on Artiﬁcial

Intelligence. Ethics guidelines for trustworthy AI,

page 6.

Holzinger, A., Saranti, A., Molnar, C., Biecek, P., and

Samek, W. (2020). Explainable AI methods-a brief

overview. In International Workshop on Extending

Explainable AI Beyond Deep Models and Classiﬁers,

pages 13–38. Springer.

Khan, S. N., Loukil, F., Ghedira-Guegan, C., Benkhelifa,

E., and Bani-Hani, A. (2021). Blockchain smart con-

tracts: Applications, challenges, and future trends.

Peer-to-peer Networking and Applications, 14:2901–

2925.

Krishna, P. R. and Karlapalem, K. (2008). Electronic con-

tracts. IEEE Internet Computing, 12(4):60–68.

Lenat, D. and Marcus, G. (2023). Getting from generative

AI to trustworthy AI: What LLMS might learn from

cyc. arXiv preprint arXiv:2308.04445.

Liang, W., Tadesse, G. A., Ho, D., Fei-Fei, L., Zaharia, M.,

Zhang, C., and Zou, J. (2022). Advances, challenges

and opportunities in creating data for trustworthy AI.

Nature Machine Intelligence, 4(8):669–677.

Lockey, S., Gillespie, N., Holm, D., and Someh, I. A.

(2021). A review of trust in Artiﬁcial Intelligence:

Challenges, vulnerabilities and future directions.

Madir, J. (2020). Smart contracts-self-executing contracts

of the future? Int’l. In-House Counsel J., 13:1.

Mason, J. (2017). Intelligent contracts and the construction

industry. Journal of Legal Affairs and Dispute Resolu-

tion in Engineering and Construction, 9(3):04517012.

McNamara, A. (2020). Automating the chaos: Intelligent

construction contracts. In Smart Cities and Construc-

tion Technologies. IntechOpen.

McNamara, A. and Sepasgozar, S. (2018). Barriers and

drivers of intelligent contract implementation in con-

struction. Management, 143:02517006.

McNamara, A. J. and Sepasgozar, S. M. (2020). Devel-

oping a theoretical framework for intelligent contract

acceptance. Construction innovation, 20(3):421–445.

Mik, E. (2017). Smart contracts: terminology, technical

limitations and real world complexity. Law, innova-

tion and technology, 9(2):269–300.

Moore, G. A. (1991). Crossing the Chasm: Marketing

and Selling High-Tech Products to Mainstream Cus-

tomers. HarperBusiness.

Munn, L. (2023). The uselessness of AI ethics. AI and

Ethics, 3(3):869–877.

Posthumus, B., Limonard, S., Schoonhoven Van, B., and

Verhagen, P. (2012). Business modelling and adoption

methodologies state-of-the-art.

Stathis, G., Biagioni, G., Trantas, A., and van den Herik,

H. J. (2023a). Risk Visualisation for Trustworthy In-

telligent Contracts. In the Proceedings of the 21st

International Industrial Simulation Conference (ISC),

EUROSIS-ETI, pages 53–57.

Stathis, G., Trantas, A., Biagioni, G., Graaf, K. A. d., Adri-

aanse, J. A. A., and van den Herik, H. J. (2023b). De-

signing an Intelligent Contract with Communications

and Risk Data. Springer Nature: Recent Trends on

Agents and Artiﬁcial Intelligence (Submitted).

Stathis, G., Trantas, A., Biagioni, G., van den Herik, H. J.,

Custers, B., Daniele, L., and Katsigiannis, T. (2023c).

Towards a Foundation for Intelligent Contracts. In the

Proceedings of the 15th International Conference on

Agents and Artiﬁcial Intelligence (ICAART).

Stathis, G. and van den Herik, J. (2023). Ethical & Preven-

tive Legal Technology. Springer Ethics & AI Journal,

Cham, Switzerland.

von Westphalen, F. G. (2017). Contracts with big data:

The end of the traditional contract concept? In Trad-

ing Data in the Digital Economy: Legal Concepts

and Tools, pages 245–270. Nomos Verlagsgesellschaft

mbH & Co. KG.

Werbach, K. (2018). The blockchain and the new architec-

ture of trust. Mit Press.

Williams, S. (2024). Generative Contracts. Arizona State

Law Journal, Forthcoming.

Xu, F., Uszkoreit, H., Du, Y., Fan, W., Zhao, D., and Zhu, J.

(2019). Explainable AI: A brief survey on history, re-

search areas, approaches and challenges. In Natural

Language Processing and Chinese Computing: 8th

CCF International Conference, NLPCC 2019, Dun-

huang, China, October 9–14, 2019, Proceedings, Part

II 8, pages 563–574. Springer.

Zheng, Z., Xie, S., Dai, H.-N., Chen, W., Chen, X., Weng,

J., and Imran, M. (2020). An overview on smart con-

tracts: Challenges, advances and platforms. Future

Generation Computer Systems, 105:475–491.

Explainable Large Language Models & iContracts

1385