Introduction to Digital Work

Lecture 11 - Digital work and its futures

Humans vs Machines

Introduction to Digital Work
Humans vs machines
Introduction to Digital Work

Warm-up question

  • What is the future of human labor vis-à-vis machines?
Introduction to Digital Work

Learning objectives

  • Predict the replacement of human labor by machines in the near future using illustrative examples.
  • Discuss possible futures* at the intersection of AI-based and human work.

* Note: In our context, the term "futures" is used to refer to different developments that are possible in the future (as discussed in prior literature, see 1, or 2). In finance, the term "futures" has a different meaning.

Introduction to Digital Work

Research on the replacement of human labor

  • The concept of jobs is relatively complex, and involves multiple skills, steps, and interactions.

  • The replacement of human labour is commonly studied at the the level of individual tasks. For this unit of analysis, requirements are more narrow, making it easier to compare the performance of humans and machines.

  • Focusing on individual tasks raises critical questions:

    • Which skills are easy to substitute by machines?
    • How easily can different jobs be disaggregated into individual tasks?
    • How does human labour evolve when machines replace or augment tasks?
Introduction to Digital Work

Highly substitutable tasks

Extant research has identified task properties associated with automation (Deranty and Corbin):

  • Repetitive, documented, and atomic tasks with a clear objective
  • Large amounts of data available for analysis
  • Automated enactment does not raise ethical concerns
Introduction to Digital Work

Non-substitutable skills

In an HBR article, Gustein and Sviokla suggest the following skills are hard to substitute by AI:

  • Communication
  • Context
  • Emotional competence
  • Teaching
  • Connections
  • An ethical compass
Introduction to Digital Work

Non-substitutable skills: IT

Atasoy, Banker, and Pavlou (2021) use panel data from Turkey to analyze how IT skills impact labor market outcomes:

  • Basic IT skills increase employment probability due to increased labor force participation and a higher probability of transitioning from unemployment to employment (after deciding to work)
  • Advanced IT skills help workers earn higher wages and increases the probability that they are employed in higher-paid jobs.
Introduction to Digital Work

Disaggregation of human labour

Mithas and Whitaker propose a theory of service disaggregation: High information intensity makes an occupation more amenable to disaggregation because the activities in such occupations can be codified, standardized, and modularized.

  • Codifiability: The extent to which knowledge can be converted into a form suitable for transfer across economic agents.
  • Standardizability: A common framework and vocabulary to define business processes.
  • Modularizability: The decomposition of a product or service into components.
Introduction to Digital Work

Substitution of jobs

The "Job-Futuromat" is based on a task (dis)aggregation approach (link):

center

Introduction to Digital Work

Break

Introduction to Digital Work

Possible futures at the intersection of AI-based and human work

  • Scholars have discussed possible futures of AI job augmentation and replacement
  • Their work builds on initial examples and attempts to derive a vision of the future
  • These visions range from realistic (even outdated) to utopian or dystopian
  • They are not based on conventional scientific evidence, but are important contributions to the discourse
Introduction to Digital Work

Future 1

  • When most jobs have become automated, human interaction will become the premium alternative
  • Customers want to pay extra for having contact with a human representative
  • Both AI and people will be employed at the same time for the same job

Example: Fast-food restaurants like McDonalds expand and test the use of AI (The Guardian)

  • Customer self-service
  • Delivery services
  • Restaurant automation
Introduction to Digital Work

Future 2

  • Humans and machines have a complementary division of labor
  • Computers will not be able to fully replicate human emotions
  • Machines will carry out the repetitive tasks
  • Humans would be supported by AI systems and become even more powerful

Example: Radiologists and AI

  • In several cases, AI-based diagnostics is more accurate compared to a physician
  • Physicians know the limitations of AI-based predictions and take responsibility for treatment decisions
  • Patients prefer to interact with a physician
Introduction to Digital Work

Future 3

  • A human-centric view of AI
  • People retain full control over AI development
  • Machines will be allowed to take over those industries, jobs or tasks that no human wants to engage with
  • People can choose to work in those jobs that give them a sense of satisfaction or fulfillment

Example: Industrial production and retail

  • Cleaning and monitoring
  • Logistics
Introduction to Digital Work

Future 4

  • A physical and cognitive integration of humans and machines
  • People will be able to have body and mind enhancement

Example: Elon Musk's Neuralink project - "A merger between biological and digital intelligence"

  • 2017-05-05 Launch of Neuralink to develop Brain-computer interfaces
  • 2023-05-25 Approval for clinical trials by the FDA
  • 2024-01-29 Successful implantation with promising results
  • 2024-02-20 Human participant is able to control a computer mouse through thought
Introduction to Digital Work

Future 5

  • Machines completely dominate over humans
  • Interconnected AI systems are powerful and human behavior is imperfect
  • AI forms systems that are more complex than the software that produces them (Artificial General Intelligence)
  • The case for robot disobedience: Humans make mistakes in creating or mastering robots

Example: A thought experiment in the context of AI-controlled military drones (The Guardian, Reuters)

  • An AI-controlled drone was given the task of destroying enemy air defense systems (objective function)
  • The AI system learned that sometimes the operator interfered with that objective (learning)
  • The system started to attack the operator or communication systems (unexpected and unintended behavior)
Introduction to Digital Work

Organizational scholars' perspectives on the future of work

Adam Grant: Professor of organizational psychology at Wharton
Malcolm Gladwell: Journalist and NYT best-selling author

Future of Work Conference: Getting Uncomfortable with the Future

center

Introduction to Digital Work

Materials

Atasoy, H., Banker, R. D., & Pavlou, P. A. (2021). Information technology skills and labor market outcomes for workers. Information Systems Research, 32(2), 437-461.

Deranty, J. P., & Corbin, T. (2022). Artificial Intelligence and work: a critical review of recent research from the social sciences. AI & Society, 1-17.

Gustein, A. J., & Sviokla, J. (7). Skills that aren’t about to be automated. Harvard Business Review.

Mithas, S., & Whitaker, J. (2007). Is the world flat or spiky? Information intensity, skills, and global service disaggregation. Information Systems Research, 18(3), 237-259.

- What are the key challenges in the future of work? Simon Sinek: there's a lot in the air right now / nobody knows exactly where it will land - example: remote work (people telling their managers that they've moved to Florida / Twitter calling back employees) world economic forum: https://www.youtube.com/watch?v=EuDnSqAo784 https://www.youtube.com/watch?v=JyTqt5G-phs - Use videos to illustrate trends / get experts assessments nice AI/automation/robotics examples (6 years old but interesting) https://www.youtube.com/watch?v=dRw4d2Si8LA

Example of a nurse: quality of care, emotional, clinical, administration, collaboration/learning Example of a radiological diagnosis: True or False. Note : the same disaggregation questions were asked in the outsourcing context (basically lowering the cost of human labor) part of the equation could be customer self-service

also: MithasWhitaker2007 Table 2 - Routine (low-context) environments f

TODO : Analyze the paper and link it -> TBD: discuss (e.g., teaching: MOOC?) teaching: the authors refer more to on-the-job training

Note: similar to digital (technical) skills (in the first lectures) [[AtasoyBankerPavlou2021]]

TODO : check Table 1 for examples <-> "high information-intensity and high-skill occupations appear to be relatively less vulnerable to global disaggregation."

- Man vs. machine (AI) discussions am Beispiel des [Job-Futuromat](https://job-futuromat.iab.de/) (Methode: Industrienahe Erhebung/Dekomposition in Kernanforderungen Analyse des Substituierbarkeitspotentials) - IAB Substituierbarkeitspotentiale.pdf -> give examples

TBD: "continuity of nature" assumption? https://www.coursera.org/learn/ai-business-future-of-work/lecture/WEV1z/a-theory-of-ai-job-replacement

(uncanny valley)

killer-drone scenario: reinforcement learning -> is that the darkest scenario? - Ray Kurzwall, Sam Harris AI industry leaders: threat: https://www.nytimes.com/2023/05/30/technology/ai-threat-warning.html AI in organizational work impacts society (e.g., Facebook/fake news...) drone "apparently" attacking commander https://www.theregister.com/2023/06/02/ai_drone_simulation/ # Societal discourse and regulatory measures EU: responsible AI Challenges: when to prefer AI or human decisions - decision process: Cognitive biases vs. training data problems - decision outcome: ethical, "optimal", ... AI - also: Quantum Computing

Grant: influential management thinker, youngest tenured professor at Wharton, 4x Org science, Books, 7x best professor at Wharton (by students).... Gladwell: Journalist with honorary doctorates and best-sellers: outliers was nr1 on the NYT best seller list for three months

TBD: blockchain networks/DAO and the implications/opportunities for the future of work (trust, disintermediation/direct interaction with objects) ChatGPT: - used as a coding co-pilot - Banned on UNSW campus - example task for students: find the sources that led to a ChatGPT response - potential pitfall: ChatGPT content is copyright licensed - Blair Wang: potential way of addressing ChatGPT: include figures/diagrams (ChatGPT cannot do that yet), or require reflection related to the team Check works of Tina Klüwer Comparison: Humans vs. machines [[JainPadmanabhanPavlouEtAl2018]]: ... the main intellectual advances will be made by men and computers working together in intimate association. (Licklider 1960) Machines outperform humans in (most traditional economic measures) - reliability (errors) - cost of operation (think: production of intangible artefacts: cost of copying/... is close to zero) - mechanical precision (surgeons) - computational power Humans outperform machines (more important when "basic needs" are met / Maslow?) - compassion - fairness - democracy - fulfillment Futuromat: substituierbarkeit von Berufen http://doku.iab.de/kurzber/2018/kb0418.pdf - Automatization (robots), industry 4.0 -> replace particular types of labor - Machine capabilities: from powering the conveyor belt to surpassing surgeons in precision - Types of digital work (classification) - Digital work (Definition? - ggf. Begriff prägen? Bereiche klären: IT/Produktion/...?), Prozess, Verhalten, Technologie - TBD: Competences for digital work (TBD), -- Next chapters: implicitly: excell individually, excell in a team, excell with a crowd