Who's in Charge? — Human Agency in AV Deployment

Businesses that master the Human-Agency Matrix use AVs as service partners to create value.

Cities, likewise, can use AVs as teammates—advancing safety, climate, and social goals.

Who is in charge?

When you step into an autonomous vehicle (AV), you relinquish lower-level controls—acceleration, braking, lane changes. Most riders barely notice and happily let go.

When an AV chooses your route, most people accept it—but any other possibilities?

• Liberty Mutual, an insurance company, would steer you along the path with the lowest crash probability;

• Climate Action Network (CAN) would choose the route with minimal CO₂ emissions;

• MassDOT would route you to ease network congestion;

• Uber would route you to maximize its profit.

• Or simply I’d rather take the scenic riverside road.

At the moment, Waymo doesn’t offer a choice. But should it? Should Waymo's route choice consider the shortest travel time for one individual AV or the lowest congestion for the whole network? Should it aim for the lowest CO2 emission or the safest path?

Even for the simple case--the classic time and money tradeoff, imagine two alternatives: A) stay in the regular lane, or B) pay $5 for the express lane and save 10 minutes. Should Waymo decide for you, infer your preference from your estimated value of time, or simply ask you?

Beyond Routing

• How about departure time: your AV might nudge, “Traffic is heavy—let’s leave 15 minutes earlier.”

• How about destination: you are heading to a steakhouse, your AV suggests, “How about a vegan bistro? You told friends you’re eating healthier.” Should it?

Three Levels of Decision-Making

Above are examples of AV decision making at three different levels:

1. Low-level decisions – acceleration, braking, lane-keeping (widely delegated to the machine).

2. Mid-level decisions – routing decision and departure time: insurers may steer you to the safest path, climate networks to the greenest, DOTs to the least-congested, and platforms to the most profitable. You want the scenic route but do you have the say?

3. High-level decisions – destination, ride-sharing vs. solo, even ownership itself: one day, your privately owned AV suggesting "You should let me go, and use Waymo instead."

How comfortable are you handing control to algorithms at each of the three levels of decision making?

AVs take agency away from us—sometimes we gladly hand it over, other times we push back.

This is NOT about the moral dilemma of trolley problems. These are practical questions AVs need to make. When shall an AV consult you? What shall your AV's default choice be?

Default is the quiet power behind every autonomous decision. The challenge is how to design it to honor the broad range of travel preferences while preserving the ease that makes autonomy appealing. Get that balance right and AVs become trusted partners.

Scenarios of operation

• Event surge: Ed Sheeran at Fenway Park jams downtown Boston. Should the AV fleet redirect riders to subway stations?

• Accident: A crash blocks I-90. Can EMS temporarily commandeer a nearby AV to assist?

• Large-scale emergency: A hurricane approaches Miami. May the DOT order all AVs to evacuate with at least four passengers each?

The human-machine balance shifts across normal traffic, events, accidents, and emergencies. Who should make decisions under each scenario--consumers, operators, car makers or authorities?

Human Agency Matrix in the AV deployment

1. Decision levels – low, mid, high, with system impacts at each level (low: eco-driving cuts emissions; mid: smarter routing boosts safety; high: centralized command enables evacuation).

2. Human vs machine balance points – fully human, fully algorithmic, and a spectrum of middle points along each level of decisions; and the heterogeneity among individual preferences for such balances.

3. Decision-making parties – individuals, operators, carmakers, public agencies, and the unavoidable tensions between them. What should be the process of reconciling such tensions?

4. Operating scenarios – normal, events, accidents, small-scale emergencies, city-wide crises. The principles and priorities differ substantially between scenarios.

Business Questions and Policy Discussions

• Should Waymo infer my willingness to pay $5 for the express lane—or ask me first?

• Clarify ownership: When I buy an AV, what exactly do I own?

• How do we design incentives so private actors want to yield control for public benefit when appropriate?

• When can MassDOT legally override all AVs for an evacuation?

Interested to receive practical insights, actionable frameworks on mobility and city policies backed by research?