Allocating the Risk of Software Failures in Automated Message Systems (Autonomous Electronic Agents)

• Jurisdiction: South Africa
• Date: 20 August 2019
• Published date: 20 August 2019
• Author: Wian Erlank
• Pages: 201-237

ALLOCATING THE RISK OF SOFTWARE

FAILURES IN AUTOMATED MESSAGE

SYSTEMS (AUTONOMOUS ELECTRONIC

AGENTS)

WIAN ERLANK*

Associate Professor in Law, North-West University

LEHLOHONOLO RAMOKANATE**

LLD Candidate, North-West University

I INTRODUCTION

A marvel of electronic commerce is an increase in the active participa-

tion of unattended computer systems in the negotiation and formation

of agreements. Computers today can, and already are, helping their

human users to make better decisions and negotiate more prof‌itable

deals.

1

Drawing from their role online, which role closely resembles that

of human agents in the real world,

2

these devices have earned the name

electronic agents.

3

The 1990s saw much talk about the so-called

information superhighway,

4

which is def‌ined as the state of affairs in

which ‘information and communication networks give access to almost

any kind of information at any time virtually anywhere in the world’.

5

The realisation of the information superhighway transformed the

*LLB (Stell) LLM LLD (Stell).

**LLB (NUL) LLM (NWU).

1

Beam & Segev, ‘Automated negotiations: A survey of the state of the art’, available at

http://www.cs.uu.nl/docs/vakken/vm/agents-negotiation.pdf, accessed on 10 August 2015;

Bainbridge, Introduction to Computer Law 4 ed (Longman 2000) 225.

2

On the comparison of electronic and human agents, see generally Wettig & Zehendner, ‘A

legal analysis of human and electronic agents’ (2004) 12(2) Artificial Intelligence and Law

111–135; Winn, ‘Emerging issues of electronic contracting, technical standards and law

reform’ (2002–2003) 7(1) Uniform Law Review 701, making the point that the main aim

behind automated tools of communication is not only to replace low level clerical employees,

but middle managers and sales representatives also; Wong, ‘The emerging law of electronic

agents: E-commerce and beyond . . .’(1999–2000) 33(1) Suffolk University Law Review 83 at

87.

3

For other names used, see Middlebrook & Muller, ‘Thoughts on bots: The emerging law

of electronic agents’ (2000) 56(1) The Business Lawyer 341 at 342. In this work, the phrases

‘electronic agents’ and ‘automated message systems’, unless otherwise indicated, shall be used

synonymously.

4

Perritt, Law and the Information Superhighway (John Wiley & Sons, Inc 1996); Smith,

Internet Law and Regulation 4 ed (Sweet and Maxwell 2007) 4.

5

Eiselen, ‘The Electronic Data Interchange Agreement’ (1995) 7(1) SA Merc LJ 1.

201

(2016) 28 SA Merc LJ 201

© Juta and Company (Pty) Ltd

internet from a simple medium of communication into a virtual

marketplace. This virtual market, also known as an information

economy,

6

is f‌illed to the brink with information about sellers, buyers,

goods, services and their prices. No human can personally search,

collect, analyse and use all this information without a waste of time,

transactional costs and prof‌its.

7

As Nimmer eloquently explains

‘[h]uman choices are a potentially inaccurate, always slower element of a

transaction’.

8

For this reason, as electronic commerce increasingly

becomes attractive for business, human involvement will not only

become undesirable, it will be actively discouraged. To conquer the

obstacles of the information economy, businesses are increasingly

resorting to automated message systems. With automated message

systems, complex transactions, being transactions that involve extensive

negotiations and comparisons, can be performed faster and at much

lower costs than by human agents.

9

Despite the magnitude of its importance, and its relevance to modern

contract law, the issue of the allocation of risk for system failures in

automated transactions has arrested very little attention in academic

commentary.

10

Automated message systems are composed of computer

hardware and software. It is the software which, with the necessary

support of hardware, is able to interact and negotiate with other

electronic agents and humans. Like all other software, the software

which runs automated message systems can malfunction. As a result of

that malfunction, automated message systems can ‘transmit an offer or

acceptance that was unintended, unforeseen or unauthorized by the

person on whose behalf the electronic agent was operating’.

11

When the

malfunction of an automated message system leads to a contract, one

6

Kephart et al, ‘Dynamic pricing by software agents’ (2000) 32(1) Computer Networks 731.

7

Turban et al, Decision Support and Business Intelligence Systems 9 ed (Prentice Hall 2011)

618–619, mentioning that the real crisis of information overload began with the advent of the

internet, and that the amount of information collected by enterprises doubles on a yearly

basis. As a result of this, it is not possible for users to personally go through all this information

without the assistance of intelligent agents.

8

Nimmer, ‘Electronic Contracting: Legal issues’ (1996) 14(2) John Marshall Journal of

Computer and Information Law 211 at 212.

9

Sorge, ‘Conclusion of contracts by electronic agents’ in International Conference on

Artificial Intelligence Law (05–11 June 2005, Bologna, Italy) 210.

10

For instance, some of the leading South African legal sources on electronic commerce

law do not in any way cover this issue in their discussion of automated transactions, see Buys

& Cronjé (eds), Cyberlaw@SA II: The Law of the Internet in South Africa 2 ed (Van Schaik

Publishers 2004) 94–96; Papadopoulos & Snail (eds), Cyberlaw@SA III: The Law of the

Internet in South Africa 3 ed (Van Schaik Publishers 2012) 55–56; Van der Merwe Computers

and the Law 2 ed (Juta & Co 2000) 241–274.

11

Kerr, ‘Ensuring the success of contract formation in agent-mediated electronic com-

merce’ (2001) 1 Electronic Commerce Research 183 at 195; Kerr, ‘Spirits in the material world:

(2016) 28 SA MERC LJ202

© Juta and Company (Pty) Ltd

way of dealing with the matter would be to identify how risk is to be

allocated between the programmer, the user of the electronic agent and

the third party who, innocently or otherwise, concluded an agreement

with the electronic agent.

12

Software developers are, however, famously

known for including extensive exemption clauses in computer con-

tracts

13

in order to limit their liability for defective software.

14

The

general attitude of courts is also skewed in favour of not holding

software manufacturers in breach of contract where defective software

has been delivered.

15

Because the liability of the manufacturer for

defective software is an independent topic, mired with its own peculiar

diff‌iculties, we shall limit the discussion of this paper to the attribution

of risk as between users of software and third parties.

The question of the allocation of risk for system failures is interesting

because, although the allocation of risk for faulty communications is a

topic well known in contract law,

16

the use of automated message

systems raises novel issues. With older technologies, especially telegrams

and faxes, messages were authored by humans. If a message was garbled

on transmission, so that it reaches the other party stating different terms

all together, a court had the luxury of choosing to enforce the contract

on the original telegram or the garbled version — that is of course if the

court was convinced that there was a valid contract despite the tele-

graphic error.

17

In automated transactions, messages are authored by

Intelligent agents as intermediaries in electronic commerce’ (1999) 22(2) Dalhousie Law

Journal 189 at 224.

12

Weitzenboeck, ‘Electronic agents and the formation of contracts’ (2001) 9(3) Interna-

tional Journal of Law and Information Technology 204 at 232.

13

End User Licence Agreements (EULAs).

14

Lloyd, Information Technology Law 3 ed (Butterworths 2000) 518. Exclusion clauses can,

of course, be overcome if they are unfair or unreasonable, see Watford Electronics Ltd v

Sanderson CFL Ltd [2001] 1 All ER (Comm) 696. For a discussion of this case, see Atkins,

‘Unfair terms in computer contracts’ in 20th British and Irish Law, Education and Technology

Association Conference (April, 2005, Queen’s University of Belfast) 8. These clauses can also be

nullif‌ied by courts construing them against software developers, see St Albans City and District

Council v International Computers Ltd [1995] FSR 686. For a discussion of this case, see

Bainbridge, (Longman 2000) 201–203.

15

Saphena Computing v Allied Collection Agencies Ltd [1995] FSR 616 at 652. For a

discussion of this and other relevant cases, see Atkins, ‘Software plus law equals bugs galore!’

(2013) 19(2) European Journal of Current Legal Issues, available at http://ojs.qub.ac.uk/

index.php/webjcli/article/view/250, accessed on 10 August 2015; Lloyd, (Butterworths 2000)

508–511.

16

For a thoughtful discussion of the law as it relates to older technologies, particularly the

telegram, see generally Ewing, ‘The effecton the formation of contract of a telegraphic mistake

in an offer’ (1933) 3(2) Idaho Law Journal 107–125; Kahn, ‘Some mysteries of offer and

acceptance’ (1955) 72(3) SALJ 263–266; Francis, ‘Two aspects of contracting by telegram’

(1967) 84 SALJ 278–294.

17

Christie, The Law of Contract 4 ed (Butterworths 2001) 85.

ALLOCATING THE RISK OF SOFTWARE FAILURES 203

© Juta and Company (Pty) Ltd