Towards Acceptance of Automated Driving Systems

Samantha Jamson

, Konstantinos Risvas

, Roi Naveiro

, David R

ıos Insua

Konstantinos Moustakas

, Mikolaj Kruszewski

, Aleksandra Rodak

and Alessandro Barisone

Institute for Transport Studies, University of Leeds, U.K.

Electrical and Computer Engineering Department, University of Patras, Greece

Institute of Mathematical Sciences (ICMAT-CSIC), Madrid, Spain

Motor Transport Institute, Warsaw, Poland

algoWatt S.p.A, Italy

{mikolaj.kruszewski, aleksandra.rodak}@its.waw.pl, alessandro.barisone@algowatt.com

Keywords:

Automated Driving Systems, Trust and Acceptance, Request to Intervene, Decision Support, Human Machine

Interfaces.

Abstract:

The acceptance of Automated Driving Systems is of key importance since it will determine whether they

will actually be used. This presentation describes contributions in this broad area from the perspective of the

Trustonomy project with a focus on ethical decision support, human machine interfaces and trust assessment,

aimed at enhancing the experience of drivers and passengers in such vehicles.

1 MOTIVATION

Automated driving systems (ADS) are poised to con-

stitute a major technological innovation reshaping

transportation as we know it. Recent breakthroughs

in Artiﬁcial Intelligence (AI), coupled with advances

in computational hardware, have had a revolution-

ary effect on ADS allowing cutting-edge control al-

gorithms to be executed in real time. Despite these

advances, it is generally recognised that ADS tech-

nology will not be widely deployed in the immediate

future: its incorporation onto global roadways will be

a gradual process (Mahmassani, 2016). Combined

with the electriﬁcation of vehicles and a change in

the concept of car ownership, ADS would conform

a future in which we would expect fewer accidents,

less pollution, less wasted travel time, and increased

traveling possibilities for many collectives, including

the elderly (Burns and Shulgan, 2019). This paper

presents contributions from the Trustonomy project

(https://h2020-trustonomy.eu/) in the areas of

ethical decision support, human machine interfaces

and human factors geared towards increasing trust and

acceptance in ADS so as to accelerate their adoption.

2 CONTEXT

A gradual progression from manned vehicles (MV)

to ADS is widely expected. Its stages are shown in

the SAE six-level driving automation taxonomy, with

level 0 describing vehicles with no automated capac-

ity, and levels 1 through 5 representing vehicles with

increasing automated features culminating in fully au-

tomated, level-5 ADS. Over the last decade, manufac-

turers have begun to produce vehicles of higher au-

tomation levels, in particular level-1 and -2. However,

many crucial limitations related to ADS safety and

operational robustness will likely restrict automated

vehicles on roads to, at best, levels 3 and 4 over the

next decade.

Figure 1 integrates relevant impacts of ADS

through a diagram illustrating the interconnected na-

ture of numerous factors related to their adoption.

Dotted nodes refer to positive impacts; light grey

nodes, to negative ones; white nodes, to other im-

pacts, not necessarily positive or negative, that need

to be addressed. Dark grey nodes refer to contextual

factors (e.g., trust in ADS) that might have a major

inﬂuence on the massive deployment of ADS.

Many factors in Figure 1 can be indeed read-

ily assumed as positive. For example, provided that

the driving error of human operators exceeds that

232

Jamson, S., Risvas, K., Naveiro, R., Insua, D., Moustakas, K., Kruszewski, M., Rodak, A. and Barisone, A.

Towards Acceptance of Automated Driving Systems.

DOI: 10.5220/0010721300003060

In Proceedings of the 5th International Conference on Computer-Human Interaction Research and Applications (CHIRA 2021), pages 232-239

ISBN: 978-989-758-538-8; ISSN: 2184-3244

Figure 1: ADS Impacts on Society. Adapted from (Caballero et al., 2021b).

of ADS technology, roadway safety can be expected

to increase. However, other impacts are perceived

negatively: although some professions may be re-

envisioned using ADS, others may be at risk, like e.g.,

taxi drivers, as a consequence of the availability of

competitive automated taxis. Similarly, the elimina-

tion of human error does not imply the elimination of

machine error. As a consequence, third parties manu-

facturing ADS safety systems will face greater vulner-

ability to liability lawsuits and reputation risk. Other

impacts associated with the massive adoption of ADS

will require a major re-deﬁnition of the current status

quo, without necessarily having positive or negative

connotations. For example, driver training will need

to evolve.

Despite the major beneﬁts that massive adoption

of ADS will bring, a core issue with these new tech-

nologies is the need to build acceptance and trust in

society to facilitate ADS adoption. This is the core

objective of the Trustonomy project which drives our

discussion here. Trustonomy involves 16 partners

across Europe covering 4 pilots in Poland/Finland,

UK, Italy and France. It is framed according to six

pillars related with: (1) driver state monitoring (DSM)

systems assessment; (2) curricula for driver training

in ADS; (3) driver intervention performance assess-

ment (DIPA); (4) ethical automated decision support

framework, covering liability and risk assessment; (5)

Human Machine Interfaces (HMI) assessment; and

(6) measuring trust and acceptance. The related nodes

in Figure 1 are marked with a red rectangle. For space

reasons, we focus on the last three topics.

3 ETHICAL DECISION SUPPORT

FOR ADS

For reasons outlined above, level -3 and -4 vehicles

are the primary focus of Trustonomy. Such ADS

require human intervention when operating outside

of their speciﬁed operational design domain (ODD).

That input is solicited through an HMI via a request to

intervene (RtI) operation, which is core in the project.

Until level-5 vehicles predominantly populate global

roadways, RtI decisions and their management will

remain a crucial, safety-related issue. Relatively few

studies focus on the RtI management. This section

sketches a solution.

Consider a level-3 or -4 ADS for which several

driving modes are available (typically, automated,

manual and emergency). A decision-analytic frame-

work is utilized to manage the RtI operations as in

Figure 2. The main workﬂow can be summarized

as follows. At each time t, the environmental and

DSM systems observe both the state of the environ-

ment and the driver. Based on past observations, a

forecast of future states is produced and the planned

trajectory is updated. If the forecasts indicate that it

is very likely that the vehicle will be outside its ODD

limits in the near future, an RtI should be executed.

If the RtI is accepted, the intervention will be as-

sessed via a DIPA. This approach emphasizes a man-

agement by exception principle (West and Harrison,

2006) wherein a group of models is used for infer-

ence, prediction and decision support under standard

Towards Acceptance of Automated Driving Systems

233

Figure 2: Decision support for ADS.

driving circumstances until an exception arises that

triggers an RtI. The approach incorporates warnings

to raise driver awareness which can be modulated in

two directions: (1) several alert levels (e.g., warning

and critical) can be introduced; (2) if an alert must be

issued repeatedly, the HMI system can successively

amplify the alert (e.g., increase volume of subsequent

warnings).

To integrate and update information from key

sources in a coherent manner, a Bayesian approach

is utilized based upon observed system behavior. D

will designate the data set available up to time t in-

corporated from the ADS sensors. Typically, most

ADS maneuvers, tasks and local trajectory planning

are scheduled a few steps ahead, e.g., k = 10 time in-

tervals of 0.5 seconds, depending on driving condi-

tions but at a minimum covering the driver’s reaction

time plus some safety buffer.

The key modules incorporated in our architecture

and utilized to manage transitions between driving

modes are: (1) an operational design domain moni-

toring system, (2) environment and DSM systems, (3)

a trajectory planning system, (4) a system for driving

mode assessment and ﬁnally (5) a module for DIPA

(details can be found in (Insua et al., 2021)). The core

system periodically issues predictive risk assessments

based on compliance with the ODD. If the predic-

tive probability of exceeding the ODD is sufﬁciently

large, the ADS alerts the driver, and assesses the au-

tomated and manual driving modes. If automated is

preferable, as the ADS is critically approaching its

design limits, the system should enter the emergency

mode and issue the appropriate alert. If manual mode

is preferred, an RtI is issued to the driver through the

HMI followed by a DIPA. If the driver performs too

poorly, as assessed from the DIPA model, the driver

is perceived not to be in good condition and the emer-

gency mode is triggered. Otherwise, the driver takes

back control until further notice.

Algorithm 1 summarises the complete ADS man-

agement procedure. There are three modes AUTO,

MANUAL, EMERG. Commands on the same line are

processed in parallel. The variables ψ

and ψ

desig-

nate the assessment (k steps ahead) of the AUTON

and MANUAL modes, respectively.

Algorithm 1: ADS controller.

Input: Priors for ODD, environment, driver state. Utility

function

Output: Trajectory from ORIGIN to DESTINY (and

implementation of commands when in AUTO or

EMERG modes).

while DESTINY not reached do

Read internal sensors. Read external sensors.

Forecast Environment k steps ahead. Forecast driver

..... state k steps ahead. Compute trajectory.

Assess driving modes (ψ

, AUTO; ψ

, MANUAL).

..... Issue WARNINGS.

Manage from DRIVING MODE. If DIPA pending,

.....resolve

end while

Our driving mode management framework is

rooted in statistical decision theory, (French and In-

sua, 2000). Therefore, the selected utility function

to assess driving modes and inform ADS’ decisions

is critical. One of the most contentious topics in

ADS research relates to their decision making in po-

tentially fatal situations, particularly the ethics as-

sociated with their automated decision making. As

is generally the case with revolutionary technolo-

gies, the widespread adoption of ADSs is accom-

panied by numerous moral uncertainties. Unfortu-

nately, in addressing these dilemmas, decision makers

are forced to grapple with unenviable ethical quan-

daries. Notably, (Awad et al., 2018) developed an on-

line experimental platform called the Moral Machine

wherein users repeatedly resolve trolley problems to

gain insight into societal ethical preferences. Alter-

natively, we propose using a generic multi-attribute

utility model for ADS management that would allow

designers, owners and policymakers to tailor ADS be-

havior according to their own ethical position. The

selected objectives, attributes and structure of the

multi-attribute utility determine the ethical perspec-

tive adopted and can accommodate multiple ethical

viewpoints as in (Keeney, 1984). Managing ADS de-

cisions in this way allows for ethical, operational, and

regulatory trade-offs to be developed and studied in a

computationally tractable manner. Moreover, this re-

search furthers the collective study of ADS ethics and

provides a means, in the far-term, to inform regula-

tion. Conversely, the near-term aim of this research

is the provision of decision-analytic support for ADS

design and operations. As these cover both short- and

long-horizon decisions in a dynamic environment, the

same set of objectives are considered in multiple de-

cision making contexts.

SUaaVE 2021 - Special Session on Research Trends to Enhance the Acceptance of Automated Vehicles

234

The ethics associated with a multi-attribute util-

ity function are primarily determined by (1) the pref-

erence model’s functional form, (2) the selected ob-

jectives, (3) the associated attributes, and (4) the ob-

jective weights utilized. The proposed objectives and

attributes are presented in Table 1, where we have

included natural, constructed and proxy attributes in

the standard decision analytic description. The pref-

erence model’s functional form follows the principles

of normative decision theory. As it is natural that the

preferences of most stakeholders will satisfy mutual

preference and utility independence (Keeney et al.,

1993), we propose using multiplicative multi-attribute

utility functions, see (Caballero et al., 2021a).

An interesting fact of managing ADS decisions

under a normative decision theoretic approach is that

the effect of objective weights on liability can be

readily observed, especially if weights are tailorable

by different stakeholders. Consider the following.

Should we assume that a normative framework is uti-

lized to model ADS preferences, then it is highly

likely that a government regulator would desire to

limit the range of objective weights speciﬁed in ADS

operations. Responding to this regulation, an ADS

manufacturer is likely to sell its vehicles with recom-

mended baseline weight settings, and it is conceiv-

able that the ADS operator would be empowered to

tailor these to their needs. However, by adapting the

weights, the operator may incur more liability, espe-

cially if the ADS’s algorithms are modiﬁed beyond

their legal limits. Indeed, a main advantage of the pro-

posed multi-attribute framework is that it sheds trans-

parency on the decision making process taken during

the design of ADSs. Utilizing this proposal, regu-

lators can undertake in-depth simulations of various

conﬁgurations until they arrive at socially acceptable

results. Such conﬁgurations can be mandated by law

or recommended as industry standards.

As an example, we illustrate how this framework

can be leveraged for liability concerns. Assume a reg-

ulator has set some safety criteria that must be met by

any ADS system to operate on public infrastructures.

The safety criteria does not distinguish between indi-

viduals inside and outside the vehicle (i.e., both are

equally weighted) and is expressed e.g. as follows:

Mean plus two standard deviations of number of in-

juries and fatalities per X kms should not be greater

than 1.4 and 0.25, respectively. From this criteria, the

regulator wishes to determine a recommended indus-

try standard for the ADS objective weights.

Simulating ADS operations, different values for

the weights of each component of the multi-attribute

utility function can be analyzed, and the regulator can

determine whether they met the criteria. Suppose the

(a) Inside safety weight of 0.1

(b) Inside safety weight of 0.7

Figure 3: Average number of injuries and fatalities vs. trip

duration weight.

chosen objectives from Table 1 are trip duration and

inside and outside safety. Figure 3a shows that if

the inside safety weight is ﬁxed at 0.1, trip duration

weights greater than or equal to 0.2 do not meet the

regulator’s criteria. Trip duration weights below 0.2

could be further explored in order to identify the max-

imum weight fulﬁlling the safety constraints. In this

particular case, a trip duration weight of 0.1 is admis-

sible.

Having completed their analysis, the regulator sets

this weight combination as a standard (0.1, 0.1, and

0.8). However, an auto manufacturer determines it

can gain market share by allocating more weight

to inside safety while maintaining the trip duration

weight constant, thereby decreasing the outside safety

weight. Concretely, suppose that an inside safety

weight of 0.7 is selected. In this scenario, if an in-

jury or fatality occurs, a natural question would be

whether or not the auto manufacturer is liable. To

make this determination, a simulation of the partic-

ular weight conﬁguration could be undertaken to as-

certain whether the results fall within the prescribed

safety bounds. Figure 3b shows that, in this particular

example, the selected weights do not meet the regula-

tor’s criteria and thus the manufacturer could reason-

ably be deemed liable.

As mentioned, the RtI would be communicated to

the driver through an HMI. Multiple authors have ex-

amined the effect of HMIs on RtIs e.g. (Walch et al.,

2015), (Eriksson and Stanton, 2017). We sketch now

Towards Acceptance of Automated Driving Systems

235

Table 1: General objectives and attributes for ADS management.

Objective

Natural

attribute

Constructed

attribute

Proxy

attribute

Min. fuel consumption Monetary

units

Min. trip duration Temporal

units

Monetary

units

Min. driver/passenger discomfort Yes ADS movement

Min. injuries of individuals inside (outside) ADS Number of

injuries

Yes No. in hospital

Min. fatalities of individuals inside (outside) ADS Number of

fatalities

VSL

Yes

Max. respect for inside (outside) ADS Probability of

death/injury

VSI

Yes

Min. damage to ADS Monetary

units

Min. infrastructure damage Monetary

units

Min. environmental impact (global/local) Monetary

units

Emissions

Min. harm to manufacturer reputation Yes Media salience

Min. harm to societal perceptions Yes Media salience

how we assess HMIs from the Trustonomy perspec-

tive.

4 HUMAN MACHINE

INTERFACES

HMIs are fundamental components of an automated

vehicle design and the main channels of information

between the vehicle and the driver. Since all automa-

tion levels between level -1 and -4 require at least sel-

dom driver interaction, the development of advanced

HMI designs is pivotal for establishing a smooth in-

teraction between the driver and the ADS. In particu-

lar, this is essential in RtI scenarios. The notiﬁcations

and information conveyed to the user should be clear

and understandable, and indicative of the occurring

driving situation. Thus, high quality HMIs capable

of generating the appropriate information are of great

importance in automated driving.

Road safety issues and the limited amount of in-

formation conveyed to the driver directed the focus

of automotive industry in developing human-centered

HMI design and assessment approaches. Many guide-

lines regarding HMI quality and assessment crite-

ria have been proposed e.g., (Naujoks et al., 2019),

(Carsten and Martens, 2019). Three fundamen-

tal features are usability, distraction and acceptance

(Franc¸ois et al., 2017). Towards this direction, the

vision of the Trustonomy project is to develop an

HMI assessment framework considering driver per-

formance, trust and acceptance.

The proposed assessment framework consists of

a time-based assessment approach, as well as an in-

novative in silico ergonomics evaluation. Validation

of the framework compels evaluating HMIs in suit-

able automotive environments (real ADS or simula-

tors) by a representative population sample with di-

verse characteristics (e.g., age, gender, driving ex-

perience) to achieve statistical signiﬁcance. Driver

and vehicle data recorded during interactions between

the participants and the available HMI designs in the

project pilots will be used as input for both frame-

work modules. The time-based module aims to eval-

uate the impact of HMIs on driver performance and

the proposed methodology includes a collection of

subjective and objective measurements. Each trial

starts with an adaptation driving scenario and, at the

end of it, the participant completes a simulation sick-

ness questionnaire. Afterwards, the actual driving

scenario initiates including secondary tasks and RtIs,

signaled by different HMI modalities. During this

time period objective measures are collected such as

the time response to the HMI stimuli. At the end

of the driving scenario, a subjective assessment fol-

lows materialized through comprehension (question-

naire designed based on ISO9186), usability (tailored

questionnaires) and workload (modiﬁed NASA-TLX)

SUaaVE 2021 - Special Session on Research Trends to Enhance the Acceptance of Automated Vehicles

236

tests. Moreover, the inﬂuence of secondary visual-

manual tasks will be assessed using the surrogate ref-

erence task (SURT).

On the other hand, the ergonomics module of

the HMI assessment framework undertakes a more

technical approach utilizing software tools to inves-

tigate the impact of the driver-HMI interactions on

the driver’s musculoskeletal system. A biomechan-

ical analysis is conducted using the OpenSim soft-

ware and appropriate musculoskeletal models that are

scaled to capture the anthropometric data of differ-

ent population classes. The pipeline requires mo-

tion capture (MoCap) data that are recorded during

the interactions between the driver and the HMI de-

sign. The ergonomics method distinguishes between

static and dynamic posture analysis. The former aims

to assess the installation position of distinct HMI de-

sign elements and the level of discomfort imposed to

the driver by estimating standard ergonomic indices

such as RULA, LUBA and NERPA. The latter evalu-

ates the entire set of interactions through the dynamic

evolution of ergonomics indices such as joint energy

and power, mean joint torque, angular impulse, and

range of motion (Kaklanis et al., 2013) (Risvas et al.,

2020). The results are displayed through heatmaps

and detailed plots and provide means of comparison

between different HMI designs.

The two proposed approaches work separately, yet

in a supplementary way, to provide assessment of the

various physical and cognitive factors that affect the

interactions between the driver and HMI design in

an automotive environment. The overall proposed

framework can be utilized by vehicle manufacturers

to improve their HMI designs. Moreover, the pro-

posed tools can be applied in simulator ADS environ-

ments, thus assisting in development of HMI proto-

types with human-centered high quality standards.

5 TRUST AND ACCEPTANCE

The concept of trust is fundamental to many of our

everyday interactions, and is especially critical when

there is uncertainty or incomplete information (Swan

and Nolan, 1985). The responses that humans demon-

strate towards technology are quite similar in some

respects to responses to other humans (Reeves and

Nass, 1996); we possibly use habits we have already

formed in interpersonal situations. Trust has been

identiﬁed as a key factor inﬂuencing acceptance and

reliance on ADS, and in particular in determining the

willingness of a human operator to rely on automa-

tion in situations of uncertainty (Lee and See, 2004).

The full safety and economic potential of ADSs will

not be reached if drivers do not accept or use them in

an appropriate way. Hence, understanding what fac-

tors associate with trust in ADSs is important in un-

derstanding and improving human interactions with

them. An individual’s personality, previous experi-

ence and the context in which they are driving are

factors that certainly inﬂuence trust in ADSs (Hoff

and Bashir, 2015). In addition, perceived ease of use,

usefulness, safety and privacy risks also mediate trust

(Zhang et al., 2019).

Within Trustonomy, and as a precursor to a driv-

ing simulator study, an online survey was carried out

to explore the “mood music” regarding trust in ADSs.

The general public may have heard media reports of

ADSs being involved in safety-critical incidents. Do

drivers have varying amounts of trust in ADSs as ex-

pressed by proxy measures such as willingness to re-

move their hands and feet from the controls, as well

as engage in tasks which direct their gaze away from

the road scene. And are there any general conclusions

that can be drawn with regards the residents of differ-

ent countries – are some more inclined to trust ADSs

in various situations than others?

Approximately 800 participants (drivers) were re-

cruited to take part in an online survey in the UK,

France, Italy and Poland. As well as ascertain de-

mographics, attitudes to automation and personal-

ity, respondents rated their trust in ADSs in eleven

road environments which varied by road type (ur-

ban/rural/motorway), complexity (links, curves or in-

tersections) and road users (present or absent). Re-

spondents were asked to “Imagine you are in the driv-

ing seat of an ADS and the ADS is in control. Be-

low are a number of driving scenarios and for each,

we would like to know how likely you would be to

take your hands off the steering wheel, rest your feet

away from the pedals, or take your eyes off the road

ahead for longer than you normally do e.g. a couple of

seconds or more.” An 11-point Likert-type scale from

0=Not at All Likely to 10=Extremely Likely was pre-

sented and levels of trust were compared across gen-

der, age, country and road scenario, see Figure 4.

Overall, respondents reported that they would re-

move their hands and feet from the controls of the

vehicle more readily than take their eyes off the road,

with males being more trusting than females. UK cit-

izens were least likely to disengage and Italians the

most likely. France and Poland were similar to each

other. As might be expected, drivers were less likely

to disengage on urban roads, compared to rural and

motorways and also when the road was complex in

design or had other road users present.

With the caveat of this being a self-report survey,

we tentatively suggest there may be between-country

Towards Acceptance of Automated Driving Systems

237

Figure 4: Sample of environments presented in the survey.

differences with regards trust, as gauged by proxy

measures of disengagement. In addition, there may

be (currently justiﬁed) reticence in trusting ADSs in

conditions where a human operator is perceived to

be performing an important monitoring and decision-

making role, such as in complex road layouts or

where other road users are present.

6 DISCUSSION

We have introduced some of the concepts that, from

an interdisciplinary perspective, are being developed

within the Trustonomy project to increase acceptance

of ADS. A framework to support ethical decisions

concerning RtIs was sketched; the entailed decisions

need to be properly communicated through appropri-

ate HMIs, demanding adequate assessment; these, in

turn, should increase trust and acceptance in ADS,

favouring their adoption and their entailed beneﬁts.

ACKNOWLEDGEMENTS

Work performed in the context of the Trustonomy

project, which has received funding from the Euro-

pean Community’s Horizon 2020 research and inno-

vation programme under grant agreement No 815003.

REFERENCES

Awad, E., Dsouza, S., Kim, R., Schulz, J., Henrich, J.,

Shariff, A., Bonnefon, Jean-Franc¸ois, R., and Iyad

(2018). The moral machine experiment. Nature,

563(7729):59–64.

Burns, L. and Shulgan, C. (2019). Autonomy: The Quest to

Build the Driverless Car—And How It Will Reshape

Our World. ECCO.

Caballero, W. N., Insua, D. R., and Naveiro, R. (2021a).

Ethical and operational preferences automated driving

systems. Tech. Rep.

Caballero, W. N., Rios Insua, D., and Banks, D. (2021b).

Decision support issues in automated driving systems.

Int. Trans. Oper. Res.

Carsten, O. and Martens, M. (2019). How can humans un-

derstand their automated cars? hmi principles, prob-

lems and solutions. Cognition, Technology & Work,

21.

Eriksson, A. and Stanton, N. A. (2017). Takeover time in

highly automated vehicles: noncritical transitions to

and from manual control. Human factors, 59(4):689–

705.

Franc¸ois, M., Osiurak, F., Fort, A., Crave, P., and Navarro,

J. (2017). Automotive hmi design and participa-

tory user involvement: review and perspectives. Er-

gonomics, 60(4):541–552. PMID: 27167154.

French, S. and Insua, D. R. (2000). Statistical Decision

Theory. Edward Arnold.

Hoff, K. A. and Bashir, M. (2015). Trust in automation: In-

tegrating empirical evidence on factors that inﬂuence

trust. Human Factors, 57(3):407–434.

Insua, D. R., Caballero, W. N., and Naveiro, R. (2021).

Managing driving modes in automated driving sys-

tems. arXiv preprint arXiv:2107.00280.

Kaklanis, N., Moschonas, P., Moustakas, K., and Tzovaras,

D. (2013). Virtual user models for the elderly and dis-

abled for automatic simulated accessibility and ergon-

omy evaluation of designs. Universal Access in the

Information Society, 12.

Keeney, R. L. (1984). Ethics, decision analysis, and public

risk. Risk Analysis, 4(2):117–129.

Keeney, R. L., Raiffa, H., and Meyer, R. F. (1993). Deci-

sions with multiple objectives: preferences and value

trade-offs. Cambridge university press.

Lee, J. and See, K. (2004). Trust in automation: Designing

for appropriate reliance. Human Factors, 46(1):50–

80.

Mahmassani, H. S. (2016). 50th anniversary invited arti-

cle—autonomous vehicles and connected vehicle sys-

tems: Flow and operations considerations. Trans-

portation Science, 50(4):1140–1162.

Naujoks, F., Wiedemann, K., Sch

omig, N., Hergeth, S., and

Keinath, A. (2019). Towards guidelines and veriﬁ-

cation methods for automated vehicle hmis. Trans-

portation Research Part F: Trafﬁc Psychology and Be-

haviour, 60:121–136.

Reeves, B. and Nass, C. (1996). The media equation: How

people treat computers, television, and new media like

real people. Cambridge University Press.

Risvas, K., Pavlou, M., Zacharaki, E., and Moustakas, K.

(2020). Biophysics-based simulation of virtual human

model interactions in 3d virtual scenes. 2020 IEEE

Conference on Virtual Reality and 3D User Interfaces

Abstracts and Workshops (VRW), pages 119–124.

Swan, J. E. and Nolan, J. J. (1985). Gaining customer trust:

a conceptual guide for the salesperson. Journal of Per-

sonal Selling Sales Management, 5(2):39–48.

SUaaVE 2021 - Special Session on Research Trends to Enhance the Acceptance of Automated Vehicles

238

Walch, M., Lange, K., Baumann, M., and Weber, M. (2015).

Autonomous driving: investigating the feasibility of

car-driver handover assistance. In Proceedings of the

7th international conference on automotive user in-

terfaces and interactive vehicular applications, pages

11–18.

West, M. and Harrison, J. (2006). Bayesian forecasting and

Dynamic Models. Springer.

Zhang, T., Tao, D., Qu, X., Zhang, X., Lin, R., and Zhang,

W. (2019). The roles of initial trust and perceived risk

in public’s acceptance of automated vehicles. Trans-

portation Research Part C, 98:207–220.

Towards Acceptance of Automated Driving Systems

239