Popular in Social Sciences

Books

City Time: On Being Sentenced to Rikers Island

Independent Scholar David A. Campbell

Culture and Society

The City as a Growth Machine: Toward a Political Economy of Place

New York University Harvey Moltoch

Economics

Gender and Economics

University of California Santa Barbara Shelly Lundberg

American Politics

Keeping the Republic

Boston College Marc Landy

Banks and Policy Organisations

Greening the financial sector

European Investment Bank Matteo Ferrazzi

British Politics

Entrepreneurial thinking: the politics and practice of policy impact

University of Sheffield Matthew Flinders

Culture and Society

Carceral Entrapments: Views from the Prison/Street Interface in India

Indian Institute of Technology Delhi Mahuya Bandyopadhya

Automation hesitancy

View more insights in Social Sciences - Sociology

View more figures from University of York

View the page of David Beer

Read the Study

David Beer discusses the concept of automation hesitancy.

Image courtesy of interviewee. December 19, 2023

Generating Workspace

Export as PDF

This video video discusses the findings of a project focused on the role of algorithms in housing decision-making, led by Alison Wallace at the University of York and funded by the Northfield Foundation. The project involves interviews with 118 individuals across the UK housing sector to understand how algorithmic decision-making influences access to various types of housing, including private rentals and social housing.

Key findings reveal a generally hesitant approach to adopting automation and algorithms in the housing sector. Contrary to the dominant narrative that algorithms are rapidly taking over decision-making processes, the study shows that different organizations implement algorithms in varying degrees, often combining human judgment with algorithmic assistance. This video video hesitancy stems from concerns about the risks associated with full automation, such as missing critical nuances that human decision-makers might catch.

The researchers identified "confidence deficits," where stakeholders are reluctant to trust algorithms entirely due to fears they might overlook important factors or lack flexibility. Despite this hesitance, there is a prevailing belief in the inevitability of increased algorithmic involvement in housing decisions.

The study emphasizes a transition in the housing sector where automation is on the rise, but certain areas still require human oversight due to the complexity of decision-making. The overall aim is to illustrate the nuanced relationship between humans and algorithms in shaping decisions that affect people's housing situations. The insights from this study are expected to be relevant across various sectors beyond housing, highlighting a broader hesitance towards automation and algorithmic processes in society.

Language:

Export as PDF

X-TIMESTAMP-MAP:LOCAL=00:00:00.000,MPEGTS=900000

00:00.115 --> 00:03.285
This, particular piece emerged from a project,

00:03.385 --> 00:06.525
on our code encounters project run by, Alison Wallace,

00:06.655 --> 00:09.045
who's also based at the University of York

00:09.505 --> 00:11.405
and funded by the Northfield Foundation.

00:12.105 --> 00:13.885
And the project's an attempt to think

00:14.025 --> 00:17.485
and to discover where algorithmic decisions shape housing

00:17.485 --> 00:20.205
choices, so access to housing.

00:20.545 --> 00:22.605
So question. This covers private rental

00:22.755 --> 00:24.085
through to social housing.

00:24.505 --> 00:26.245
Um, we cover mortgage brokers,

00:26.625 --> 00:28.645
so there's interviews being done with,

00:29.205 --> 00:32.685
118 people across the housing sector in the UK

00:33.305 --> 00:35.525
to get different perspectives from the talent through

00:35.525 --> 00:38.045
to the, the decision makers in the agents.

00:38.785 --> 00:41.845
Um, so what we, one thing we'd, we'd been looking at was,

00:41.905 --> 00:45.205
the powerful ideas about the, the role

00:45.205 --> 00:47.805
of the algorithm in making decisions about us.

00:48.745 --> 00:52.125
Um, and when we looked more closely at the housing sector

00:52.125 --> 00:54.205
through these interviews, what we found,

00:54.825 --> 00:58.045
was a different picture really, of a very,

00:58.415 --> 01:02.005
variegated take up of, of, different levels

01:02.105 --> 01:04.605
of algorithmic processing and decision making.

01:05.625 --> 01:07.885
In this case, when we talk about algorithms here,

01:07.885 --> 01:10.085
we're thinking about, bits of software code

01:10.085 --> 01:11.525
that take decisions for us

01:11.625 --> 01:13.725
or on the behalf of, of human actors.

01:14.585 --> 01:18.005
Um, so what we found was actually a very quite hesitant

01:18.365 --> 01:21.645
approach towards the implementation of automation

01:21.705 --> 01:24.965
and algorithms within the housing sector in the uk.

01:25.505 --> 01:26.965
And we wanted to tell that story

01:26.965 --> 01:28.085
because it's quite different

01:28.105 --> 01:30.685
and it builds up a different pitch to the, the dominant idea

01:30.685 --> 01:32.085
that algorithms are coming in,

01:32.395 --> 01:34.205
they're implemented straight away,

01:34.505 --> 01:36.525
and they're suddenly taking all the decisions.

01:37.195 --> 01:40.325
What we're finding instead is different levels of,

01:40.385 --> 01:43.925
of hesitancy going on in the implementation of automation

01:44.595 --> 01:47.045
that leads to different parts of the sector

01:47.105 --> 01:49.725
and different companies and organizations having different

01:49.725 --> 01:53.085
levels of algorithmic decision within their organization.

01:53.785 --> 01:55.725
So some are more algorithmic than others,

01:56.305 --> 01:59.405
but often systems combine human decision making

01:59.835 --> 02:02.885
with algorithmic input in different combinations.

02:03.065 --> 02:05.885
So automation hesitancy is important

02:05.885 --> 02:09.205
because it shapes the way decisions are made about people in

02:09.205 --> 02:12.445
the housing sector, and it also shows that it isn't

02:12.445 --> 02:14.245
as straightforward as suddenly the decisions

02:14.245 --> 02:15.325
are made by algorithms.

02:15.515 --> 02:18.605
They're made by quite complicated combinations of humans

02:18.665 --> 02:22.525
and algorithms together as a result of the kind of hesitancy

02:22.795 --> 02:25.005
that people have towards suddenly

02:25.635 --> 02:29.445
turning their company their decisions over to the algorithm

02:30.075 --> 02:31.925
they want to hold onto their expertise.

02:32.505 --> 02:35.165
Uh, they have a heightened sense of the risks that come

02:35.165 --> 02:36.165
with the algorithms,

02:36.905 --> 02:39.245
but they also see that there are possibilities there.

02:39.265 --> 02:41.285
So there's this, hesitancy,

02:41.285 --> 02:43.405
even though the algorithm continues

02:43.405 --> 02:45.245
to be implemented within the sector.

02:45.715 --> 02:46.925
What we found was there were

02:46.925 --> 02:49.525
what we've called in the paper confidence deficits.

02:49.705 --> 02:50.805
So there were, there are areas

02:50.815 --> 02:53.805
where people weren't confident in handing the decision

02:53.805 --> 02:54.845
over to the algorithm.

02:55.585 --> 02:58.205
So this, this occurred for a number of reasons. Um,

02:58.825 --> 02:59.885
Uh, one is that it,

02:59.885 --> 03:03.045
they were afraid the algorithm might miss something,

03:03.145 --> 03:06.005
or that the algorithm lacked flexibility in its decision

03:06.005 --> 03:07.085
making possibility.

03:07.185 --> 03:09.405
So it would rule out people

03:09.465 --> 03:11.645
who would be perfectly good tenants, for instance.

03:12.265 --> 03:14.485
Um, so there's the, those, those kind

03:14.485 --> 03:17.725
of confidence deficits, the gaps, the sort of,

03:18.505 --> 03:19.685
the attempt to protect

03:19.745 --> 03:22.725
or to height a heightened sense of risk that came with

03:23.285 --> 03:26.045
suddenly handing over their decisions to an automated system

03:26.675 --> 03:29.125
that was part of the reason that people became hesitant.

03:29.185 --> 03:32.085
So within the sector, there's a sense of, of risk

03:32.085 --> 03:36.125
of full automation, a sense of the, the problems

03:36.125 --> 03:39.525
that might bring, and also the things it might miss

03:40.275 --> 03:41.965
that a human being can spot.

03:42.035 --> 03:44.205
There's, these are the things that were being protected

03:44.705 --> 03:47.285
and also was being protected was a sense of,

03:47.505 --> 03:49.445
the expertise of the decision maker,

03:49.825 --> 03:50.845
within those sectors.

03:50.845 --> 03:53.725
You know, having a feel for the right tenant and the,

03:53.785 --> 03:55.725
or what would make, who would be a good tenant

03:55.745 --> 03:56.845
who would be trustworthy.

03:57.675 --> 04:00.245
This was seen as a bit more complicated than the algorithm

04:00.505 --> 04:01.885
can potentially handle.

04:02.315 --> 04:03.765
This is where the variation comes.

04:03.825 --> 04:06.165
So it can be doing different things across the sector.

04:06.165 --> 04:07.325
This is the, one of the things we wanted

04:07.325 --> 04:09.085
to highlight is this variation in

04:09.085 --> 04:11.565
where the algorithm is, is being used.

04:12.025 --> 04:14.325
Um, but one of the things it can be used to for,

04:14.325 --> 04:16.085
for instance, is to decide which tenant

04:16.145 --> 04:17.565
to go for in decisions.

04:18.065 --> 04:20.645
Um, who's the, who's the most trustworthy of these,

04:20.645 --> 04:22.885
you know, so it can help with those sorts of decisions.

04:23.345 --> 04:25.005
It can highlight potential risks

04:25.105 --> 04:27.805
or problems within, so that decisions can then be made

04:28.805 --> 04:30.365
informed by those risks.

04:30.865 --> 04:32.285
It can be to do with credit.

04:32.735 --> 04:35.085
We've also been looking at and credit referencing.

04:35.085 --> 04:38.925
It can also be to do with, referencing more generally

04:39.105 --> 04:41.205
or open banking is one of the things

04:41.205 --> 04:42.205
that we've found in the sector.

04:42.345 --> 04:45.925
So, in order to be able to be considered to be able

04:45.925 --> 04:48.525
to credit, to sign up for, from banking so

04:48.525 --> 04:51.285
that data can be seen about the individual,

04:51.345 --> 04:52.365
and their transactions.

04:52.385 --> 04:54.245
So there's different levels in which the,

04:54.385 --> 04:55.685
the algorithm is applied and it,

04:55.705 --> 04:57.845
it varies across a lot across the sector,

04:57.905 --> 04:59.845
but they're some of the types of things that are,

05:00.265 --> 05:01.365
are being implemented.

05:01.825 --> 05:03.605
We do look at it in the way that we are think,

05:03.605 --> 05:05.085
we're looking at how ideas

05:05.185 --> 05:07.845
of efficiency are embedded into the decisions

05:07.845 --> 05:08.925
that are being made about

05:08.925 --> 05:10.525
where the algorithm is implemented.

05:10.945 --> 05:14.005
So one of the arguments that you find within the sector is

05:14.005 --> 05:16.725
that the algorithm does bring a combination of speed

05:16.865 --> 05:20.125
and efficiency, so the decisions can be made more quickly.

05:20.555 --> 05:22.925
Sometimes the idea is that's more accurate as well.

05:23.345 --> 05:27.525
Um, but there also those, those quicker decisions mean

05:27.565 --> 05:29.325
that there are parts of the sector

05:29.325 --> 05:30.725
that can then shrink a little bit

05:30.745 --> 05:32.605
or become more efficient in response.

05:33.345 --> 05:37.285
Um, so, as a result of that, you do the, the notion

05:37.285 --> 05:38.845
of efficiency becomes important

05:38.845 --> 05:41.085
because it's part of how people see the sector

05:41.145 --> 05:43.685
and how they, make decisions about where

05:43.685 --> 05:47.165
to include the algorithm in their processes

05:47.225 --> 05:49.565
and what they want to speed up and what they don't.

05:50.305 --> 05:52.405
Um, and we found in some cases it was

05:52.405 --> 05:56.085
that speed up in efficiency that people were a bit,

05:56.115 --> 05:59.045
worried about that created this confidence deficits

05:59.045 --> 06:00.325
because they were thinking, well,

06:00.325 --> 06:02.285
if we don't have those people overseeing it,

06:02.675 --> 06:03.885
then there's a risk there.

06:04.145 --> 06:05.605
Or if we speed up too much,

06:05.895 --> 06:07.205
we'll miss something, make mistakes.

06:07.225 --> 06:08.285
So that's it.

06:08.305 --> 06:10.405
It was a sense of the algorithm brings that

06:10.405 --> 06:13.925
as a potential thing, but also that created a sense of risk.

06:14.545 --> 06:16.045
So what we found was that

06:16.165 --> 06:17.845
although there were these confidence deficits

06:17.845 --> 06:20.285
that led people to be a bit hesitant about bringing

06:20.495 --> 06:22.325
automation into the organization

06:22.425 --> 06:27.405
or extending its reach, we also, noticed within the data

06:27.405 --> 06:28.525
that there's a very strong sense

06:28.525 --> 06:29.885
that the automation will come,

06:29.955 --> 06:32.725
that the horizon is increasing automation.

06:33.425 --> 06:37.245
The, the horizon is like ever been more algorithmic.

06:37.825 --> 06:40.325
So although there's a sense of things that pe they,

06:40.325 --> 06:42.525
that people want to protect, that they think they need

06:42.525 --> 06:44.965
to hold onto and they need to retain a kind

06:44.965 --> 06:46.885
of oversight in the al of the algorithm,

06:47.025 --> 06:49.125
or they need to retain their expertise

06:49.305 --> 06:52.605
or their flexible decision making at the same time,

06:52.705 --> 06:54.685
we noticed that embedded in this is a sense

06:54.685 --> 06:56.965
that in the future that the al

06:56.965 --> 06:59.565
that the algorithm will be increasingly

06:59.805 --> 07:00.885
involved in the decision making.

07:01.315 --> 07:04.245
It's just then there are areas that might still be protected

07:04.475 --> 07:07.045
because they're seen to need human intervention,

07:07.905 --> 07:11.045
but that there's still a move towards perhaps not full

07:11.055 --> 07:13.885
automation, but that there's a direction of travel.

07:14.225 --> 07:16.685
So one of the things we've noticed is this is a period

07:16.975 --> 07:19.925
where the housing sector is almost talking about kind

07:19.925 --> 07:21.005
of transition and change.

07:21.195 --> 07:24.365
It's, it's, it's the, we are seeing something, the project

07:24.505 --> 07:27.045
as we've been doing it is, is identifying kind of time

07:27.065 --> 07:30.045
of transition in which people have got a sense

07:30.045 --> 07:31.525
that there's more automation

07:31.525 --> 07:33.765
and more algorithmic input to come in the,

07:34.105 --> 07:37.365
in the near future, and that they're trying to work out how

07:37.365 --> 07:38.365
to move towards that.

07:38.365 --> 07:41.045
And there's a kind of sense of inevitability really,

07:41.205 --> 07:43.805
I think in that, in some cases very near.

07:44.385 --> 07:47.205
So in some cases it's, we are in the process

07:47.465 --> 07:50.405
of making up these parts of our referencing

07:50.505 --> 07:52.765
or these parts of our form filling

07:53.865 --> 07:55.365
in incre or more automated.

07:55.385 --> 07:57.325
So it's almost, you know, this is underway.

07:57.985 --> 08:00.525
And then there are things to do with the overall with the,

08:00.545 --> 08:01.885
the, some of the bigger decisions

08:01.985 --> 08:05.245
or, the judgments around tenant, for instance,

08:05.245 --> 08:06.285
that just feel like they're,

08:06.285 --> 08:07.565
they're a bit more in the distance

08:07.565 --> 08:09.965
and they don't actually know what form they will quite take

08:10.145 --> 08:13.605
yet, but they know that the, that the, the technology

08:13.745 --> 08:16.565
and the software is coming that will facilitate that.

08:16.625 --> 08:18.245
And also the different data sources

08:18.315 --> 08:19.645
that would need to feed into it.

08:20.025 --> 08:21.605
So there's, there's a, there's a sense

08:21.605 --> 08:23.605
of what's coming in the immediate,

08:24.105 --> 08:26.805
but there's also this more distant sense of a, a bit

08:26.805 --> 08:30.605
of an unknown future that they know they're heading towards

08:31.265 --> 08:33.925
but aren't quite sure exactly what format's gonna take.

08:33.985 --> 08:37.045
And actually that's the next thing I'm working on in the

08:37.045 --> 08:39.125
project is from within the housing sector,

08:39.155 --> 08:42.125
what are their visions of the algorithmic future Exactly.

08:42.265 --> 08:44.125
And how can we understand them?

08:44.445 --> 08:47.445
'cause we want to think about how we can feed into

08:47.905 --> 08:48.965
the future of the sector.

08:49.535 --> 08:51.885
There are powerful promises attached to these types

08:51.885 --> 08:53.925
of technologies in, in a broader sense,

08:54.385 --> 08:57.325
not just in the housing sector that draw people in

08:57.325 --> 09:00.005
and elsewhere talks about how the frontiers are kind

09:00.005 --> 09:03.285
of pushed back, sort of data frontiers within these sectors

09:03.285 --> 09:04.765
get pushed back by the promises.

09:04.825 --> 09:08.325
So the sense of what it will bring if we change,

09:09.025 --> 09:11.485
is quite powerful in shaping the present.

09:11.825 --> 09:16.125
So some, although some of those distant future horizons,

09:16.345 --> 09:19.365
are a bit unknown, they're still quite powerful in the

09:19.395 --> 09:22.165
direction in which they take decisions in the present.

09:23.265 --> 09:26.005
And often these technologies have very powerful,

09:26.705 --> 09:30.165
big promises attached to them about things like efficiency

09:30.905 --> 09:35.645
or accuracy of decision, or, speed.

09:36.195 --> 09:38.125
Lots of those types of things are attached to 'em

09:38.125 --> 09:39.845
as quite powerful promises that,

09:40.275 --> 09:42.405
that just push back those frontiers and,

09:42.405 --> 09:45.005
and it'll lead to increasing automation,

09:45.005 --> 09:47.285
increasing application of the algorithm, the kind

09:47.285 --> 09:49.165
of promises of what that future will be.

09:49.785 --> 09:51.005
The main starting point

09:51.005 --> 09:54.045
for this is an early paper on the project was that we wanted

09:54.185 --> 09:58.405
to build up a, a, a more detailed picture of

09:58.405 --> 10:00.725
what was actually happening within the housing sector

10:00.725 --> 10:02.045
with regard to algorithms.

10:02.045 --> 10:04.205
So that's where we focus this paper.

10:04.745 --> 10:08.085
Um, and the, the insight on automation hesitancy

10:08.865 --> 10:11.605
we think is gonna be useful across different sectors.

10:11.625 --> 10:14.325
So although we've looked at housing here,

10:14.905 --> 10:17.525
we have a strong sense that this kind of

10:18.045 --> 10:20.845
hesitant approach towards automation where people pause

10:21.025 --> 10:25.045
or aren't quite sure or feel a sense of risk

10:25.105 --> 10:28.045
or their confidence isn't quite there, mean

10:28.045 --> 10:30.005
that there's a hesitant approach.

10:30.105 --> 10:32.805
So where we see promises of these technologies,

10:32.955 --> 10:34.765
when people actually come to apply them,

10:34.835 --> 10:36.405
they're a little bit more hesitant

10:36.465 --> 10:39.165
and they pause a little bit more in, in those applications.

10:39.425 --> 10:42.085
We think that will, that's likely to be something

10:42.085 --> 10:44.845
that will appear across different sectors beyond housing.

10:45.705 --> 10:48.765
Um, and it's also then we think that

10:48.765 --> 10:50.405
that tells us something about the kind

10:50.405 --> 10:52.645
of varied ways in which humans

10:52.745 --> 10:55.085
and machines knit together in the decisions

10:55.085 --> 10:57.605
and judgments that are made about us in everyday life.

10:58.545 --> 11:01.005
And that's the picture that we wanted to, to offer

11:01.005 --> 11:01.685
with this paper.

00:00 - This, particular piece emerged from a project, on our code encounters project run by, Alison Wallace, who's also based at the University of York and funded by the Northfield Foundation. And the project's an attempt to think and to discover where algorithmic decisions shape housing

00:17 - choices, so access to housing. So question. This covers private rental through to social housing. we cover mortgage brokers, so there's interviews being done with, 118 people across the housing sector in the UK to get different perspectives from the talent through to the, the decision makers in the agents. so what we, one thing we'd, we'd been looking at was, the powerful ideas about the, the role

00:45 - of the algorithm in making decisions about us. and when we looked more closely at the housing sector through these interviews, what we found, was a different picture really, of a very, variegated take up of, of, different levels of algorithmic processing and decision making. In this case, when we talk about algorithms here, we're thinking about, bits of software code that take decisions for us or on the behalf of, of human actors.

01:14 - so what we found was actually a very quite hesitant approach towards the implementation of automation and algorithms within the housing sector in the uk. And we wanted to tell that story because it's quite different and it builds up a different pitch to the, the dominant idea that algorithms are coming in, they're implemented straight away, and they're suddenly taking all the decisions. What we're finding instead is different levels of,

01:40 - of hesitancy going on in the implementation of automation that leads to different parts of the sector and different companies and organizations having different levels of algorithmic decision within their organization. So some are more algorithmic than others, but often systems combine human decision making with algorithmic input in different combinations. So automation hesitancy is important because it shapes the way decisions are made about people in the housing sector, and it also shows that it isn't

02:12 - as straightforward as suddenly the decisions are made by algorithms. They're made by quite complicated combinations of humans and algorithms together as a result of the kind of hesitancy that people have towards suddenly turning their company their decisions over to the algorithm they want to hold onto their expertise. they have a heightened sense of the risks that come with the algorithms, but they also see that there are possibilities there.

02:39 - So there's this, hesitancy, even though the algorithm continues to be implemented within the sector. What we found was there were what we've called in the paper confidence deficits. So there were, there are areas where people weren't confident in handing the decision over to the algorithm. So this, this occurred for a number of reasons. one is that it,

02:59 - they were afraid the algorithm might miss something, or that the algorithm lacked flexibility in its decision making possibility. So it would rule out people who would be perfectly good tenants, for instance. so there's the, those, those kind of confidence deficits, the gaps, the sort of, the attempt to protect or to height a heightened sense of risk that came with suddenly handing over their decisions to an automated system

03:26 - that was part of the reason that people became hesitant. So within the sector, there's a sense of, of risk of full automation, a sense of the, the problems that might bring, and also the things it might miss that a human being can spot. There's, these are the things that were being protected and also was being protected was a sense of, the expertise of the decision maker, within those sectors. You know, having a feel for the right tenant and the,

03:53 - or what would make, who would be a good tenant who would be trustworthy. This was seen as a bit more complicated than the algorithm can potentially handle. This is where the variation comes. So it can be doing different things across the sector. This is the, one of the things we wanted to highlight is this variation in where the algorithm is, is being used. but one of the things it can be used to for,

04:14 - for instance, is to decide which tenant to go for in decisions. who's the, who's the most trustworthy of these, you know, so it can help with those sorts of decisions. It can highlight potential risks or problems within, so that decisions can then be made informed by those risks. It can be to do with credit. We've also been looking at and credit referencing. It can also be to do with, referencing more generally

04:39 - or open banking is one of the things that we've found in the sector. So, in order to be able to be considered to be able to credit, to sign up for, from banking so that data can be seen about the individual, and their transactions. So there's different levels in which the, the algorithm is applied and it, it varies across a lot across the sector, but they're some of the types of things that are,

05:00 - are being implemented. We do look at it in the way that we are think, we're looking at how ideas of efficiency are embedded into the decisions that are being made about where the algorithm is implemented. So one of the arguments that you find within the sector is that the algorithm does bring a combination of speed and efficiency, so the decisions can be made more quickly. Sometimes the idea is that's more accurate as well.

05:23 - but there also those, those quicker decisions mean that there are parts of the sector that can then shrink a little bit or become more efficient in response. so, as a result of that, you do the, the notion of efficiency becomes important because it's part of how people see the sector and how they, make decisions about where to include the algorithm in their processes and what they want to speed up and what they don't.

05:50 - and we found in some cases it was that speed up in efficiency that people were a bit, worried about that created this confidence deficits because they were thinking, well, if we don't have those people overseeing it, then there's a risk there. Or if we speed up too much, we'll miss something, make mistakes. So that's it. It was a sense of the algorithm brings that

06:10 - as a potential thing, but also that created a sense of risk. So what we found was that although there were these confidence deficits that led people to be a bit hesitant about bringing automation into the organization or extending its reach, we also, noticed within the data that there's a very strong sense that the automation will come, that the horizon is increasing automation. The, the horizon is like ever been more algorithmic.

06:37 - So although there's a sense of things that pe they, that people want to protect, that they think they need to hold onto and they need to retain a kind of oversight in the al of the algorithm, or they need to retain their expertise or their flexible decision making at the same time, we noticed that embedded in this is a sense that in the future that the al that the algorithm will be increasingly involved in the decision making.

07:01 - It's just then there are areas that might still be protected because they're seen to need human intervention, but that there's still a move towards perhaps not full automation, but that there's a direction of travel. So one of the things we've noticed is this is a period where the housing sector is almost talking about kind of transition and change. It's, it's, it's the, we are seeing something, the project as we've been doing it is, is identifying kind of time of transition in which people have got a sense

07:30 - that there's more automation and more algorithmic input to come in the, in the near future, and that they're trying to work out how to move towards that. And there's a kind of sense of inevitability really, I think in that, in some cases very near. So in some cases it's, we are in the process of making up these parts of our referencing or these parts of our form filling in incre or more automated.

07:55 - So it's almost, you know, this is underway. And then there are things to do with the overall with the, the, some of the bigger decisions or, the judgments around tenant, for instance, that just feel like they're, they're a bit more in the distance and they don't actually know what form they will quite take yet, but they know that the, that the, the technology and the software is coming that will facilitate that. And also the different data sources

08:18 - that would need to feed into it. So there's, there's a, there's a sense of what's coming in the immediate, but there's also this more distant sense of a, a bit of an unknown future that they know they're heading towards but aren't quite sure exactly what format's gonna take. And actually that's the next thing I'm working on in the project is from within the housing sector, what are their visions of the algorithmic future Exactly. And how can we understand them?

08:44 - 'cause we want to think about how we can feed into the future of the sector. There are powerful promises attached to these types of technologies in, in a broader sense, not just in the housing sector that draw people in and elsewhere talks about how the frontiers are kind of pushed back, sort of data frontiers within these sectors get pushed back by the promises. So the sense of what it will bring if we change, is quite powerful in shaping the present.

09:11 - So some, although some of those distant future horizons, are a bit unknown, they're still quite powerful in the direction in which they take decisions in the present. And often these technologies have very powerful, big promises attached to them about things like efficiency or accuracy of decision, or, speed. Lots of those types of things are attached to 'em as quite powerful promises that, that just push back those frontiers and, and it'll lead to increasing automation,

09:45 - increasing application of the algorithm, the kind of promises of what that future will be. The main starting point for this is an early paper on the project was that we wanted to build up a, a, a more detailed picture of what was actually happening within the housing sector with regard to algorithms. So that's where we focus this paper. and the, the insight on automation hesitancy we think is gonna be useful across different sectors.

10:11 - So although we've looked at housing here, we have a strong sense that this kind of hesitant approach towards automation where people pause or aren't quite sure or feel a sense of risk or their confidence isn't quite there, mean that there's a hesitant approach. So where we see promises of these technologies, when people actually come to apply them, they're a little bit more hesitant and they pause a little bit more in, in those applications.

10:39 - We think that will, that's likely to be something that will appear across different sectors beyond housing. and it's also then we think that that tells us something about the kind of varied ways in which humans and machines knit together in the decisions and judgments that are made about us in everyday life. And that's the picture that we wanted to, to offer with this paper.

Faculti AI can make mistakes